diff --git a/Simul-Robot/c001_episode_000000/sample.json b/Simul-Robot/c001_episode_000000/sample.json
index 60c74a5625cd38a764ff657b26f878bd378a054c..8edba6f6eefbdff402765ceda4f5c574e42a07c4 100644
--- a/Simul-Robot/c001_episode_000000/sample.json
+++ b/Simul-Robot/c001_episode_000000/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "ab33b93c4890ec2b8b62c5cde28124377ebd70ce2e5d5ccb355da2b9b8ea29b0",
   "labels/objects.json": "90c12e04f30944528bf61510e16af4c7a7a22e3057a65179876d19ee1cb746f2",
   "labels/visibility_states.json": "346d655a413bf618e2d2fdc77f1fd15072aaad2728a4bec8ad302b7a680aecd3",
-  "source/cameras.json": "48d6761d6163cf27ff2e0208f7eb0fb2002b36c1c9b8f8c596d1aaf5d0e00c3e",
+  "source/cameras.json": "5b86455279665283aed35557ca529cbf9a66e7d75876a14395ac3406d4cda9b9",
   "source/caption.txt": "b0b9403d1cc417942a73a9fbb4458a890df4e6f548a441d71ae9911d55f6dc5a",
-  "source/meta.json": "24146e046a79b3331995068e968e02fee818cb14e60615f990fc283a0b901287",
-  "source/points_target.npz": "43d8b367f034a31f1270d151f7b15f58302fb9585a986a85eed2dd30007e6fbe",
-  "source/visibility.json": "11931ceb9d6fa93b6dcf71ddca0bfb35ee0d39b1884baa51ac81348876d2e3df",
+  "source/meta.json": "eff6917d5f420c9fea3fd6e015aa6034808f455b822598e7c2ea61923eb042d7",
+  "source/points_target.npz": "137e904dedde5ca378a8c77c76d8592f2293c84b82c403e2ac3b2a71fc1fb2d0",
+  "source/visibility.json": "d6f38dc8d018ad5df2659689617705ad1057115ec608ad185262fee99ce34d95",
   "tracks3d.npz": "aebcc6e6971a927498070fbb614b6b6cb8353f9761acd223e3602b7855945b0a",
   "videos/view0.mp4": "b8bbc23a9d67d405db5edac0f70734b1c1627bb0f1318a0f8359babde80df382",
   "videos/view1.mp4": "806459afd9146a156f782e7581e5ee8a441e7ca2f6805975c6a1a3bf77820ac4",
diff --git a/Simul-Robot/c001_episode_000000/source/cameras.json b/Simul-Robot/c001_episode_000000/source/cameras.json
index 03859c3d100313139409d53b1741c88ccc658437..86e0693769ea58dc22daee6068a36adbb2f78bcf 100644
--- a/Simul-Robot/c001_episode_000000/source/cameras.json
+++ b/Simul-Robot/c001_episode_000000/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_000000/source/meta.json b/Simul-Robot/c001_episode_000000/source/meta.json
index 781bfe752db041adf186e49f7ccaa813ed1d9fa7..d4e699ada5cd9e258d2ce8e7703b40600edc93f0 100644
--- a/Simul-Robot/c001_episode_000000/source/meta.json
+++ b/Simul-Robot/c001_episode_000000/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_000000/source/visibility.json b/Simul-Robot/c001_episode_000000/source/visibility.json
index 874ff7e81d9ac28766fae82925b2d08848ea00aa..b02f3844229f470f38de50076814c1d17e151f20 100644
--- a/Simul-Robot/c001_episode_000000/source/visibility.json
+++ b/Simul-Robot/c001_episode_000000/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "small_yellow_cube": [
@@ -6700,9 +6700,9 @@
       "target": "small_yellow_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -11150,5 +11150,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_000040/sample.json b/Simul-Robot/c001_episode_000040/sample.json
index 57870dd67c91a84bb76eb1c82b544242169fe2fb..0621266bc58942898ac2da8d58dee710e4585502 100644
--- a/Simul-Robot/c001_episode_000040/sample.json
+++ b/Simul-Robot/c001_episode_000040/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "2e95149cf05b152249dffe19149765943f6c1bc6e45f055d6f195af4ac5810c8",
   "labels/objects.json": "25e5239cdceae6fc9226b1627e51211c8c551ee5d3d3d2d493502533ce078cab",
   "labels/visibility_states.json": "3f5369b5bed128dbeceb2fff555f4a60cffb03ac07e303cd018916450aa5c29e",
-  "source/cameras.json": "ae2e38e66275c4a317260ac699e64c4f470b219375d8417324971adace1d3c04",
+  "source/cameras.json": "925e1bc1fb0b904fdda9d39f635f9820f777ab8b17b61b7088f0469de367314b",
   "source/caption.txt": "4e0ba0de03c5df18a2de250cdf1d44f5a11d7ba469d1d7f9e6ed5ac84698f0d3",
-  "source/meta.json": "b954be541650d8d63528bf84a5d2614333067141d8e8e48bb5657021d1fcb030",
-  "source/points_target.npz": "8562dc661899a9f747dde57b74471486115206e8e86518f5ab4976cb77c3d65f",
-  "source/visibility.json": "15e04e6fcfe0d8a77552548bdf4990e5aab22a12d75f1484ffb59915c05c50ec",
+  "source/meta.json": "c7b5b380541266c71fc736bb75190069ead8b3a6918b8680a9d96c9e1cf834d7",
+  "source/points_target.npz": "53855d7435a4f274facaf42cb0ce607db443a5f19914be9639ed5a0c84284763",
+  "source/visibility.json": "e818231ab35d73eef47094e3822b42d9b8fa18d86468fc0abd907ca8211ff2ad",
   "tracks3d.npz": "98855cc0221d7ba3adeeb051d449ec11d348cd024964af2f7a6c4fa3ae9da50c",
   "videos/view0.mp4": "11d4418dbc9b6451c95f1a8d1569915c9850f130816d2f421765cbdea7c51fea",
   "videos/view1.mp4": "413d4104a415894bfca9e0505f97d547048d8e274e9f80229aea00173678790e",
diff --git a/Simul-Robot/c001_episode_000040/source/cameras.json b/Simul-Robot/c001_episode_000040/source/cameras.json
index f9b9167522740eaaa35b03237f51f6afc120a082..9f6d9f73d6a3ea2bb758899cfada03c8bafa8acc 100644
--- a/Simul-Robot/c001_episode_000040/source/cameras.json
+++ b/Simul-Robot/c001_episode_000040/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_000040/source/meta.json b/Simul-Robot/c001_episode_000040/source/meta.json
index 8053a7dc222a848634753bc6c4c139250bb95a76..5734e8874e310fd996be19a523d9ec5a5e6e808e 100644
--- a/Simul-Robot/c001_episode_000040/source/meta.json
+++ b/Simul-Robot/c001_episode_000040/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 1.37] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 1.37] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_000040/source/visibility.json b/Simul-Robot/c001_episode_000040/source/visibility.json
index 4e1ede7833975c5a0a8677072199ce7347a09c5b..dae0751cf2910e51dc184717676a94a44181e93b 100644
--- a/Simul-Robot/c001_episode_000040/source/visibility.json
+++ b/Simul-Robot/c001_episode_000040/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "small_blue_cube": [
@@ -5893,8 +5893,8 @@
       "target": "small_blue_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 1.37] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 1.37] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9802,5 +9802,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_000180/sample.json b/Simul-Robot/c001_episode_000180/sample.json
index 41a5aec57e82e0012aec8528acc40cd1572324cc..f08a7fde405ab8b5c57697b61ddeb739190a0d6e 100644
--- a/Simul-Robot/c001_episode_000180/sample.json
+++ b/Simul-Robot/c001_episode_000180/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "c9b2a640bbcbd03f405a28e6aabfd57c47d974b4d516575ee1cdc2ea9a0d279d",
   "labels/objects.json": "6de3a8cb64ea769bc105910630f6641553ccfd5ba49176cb34db781e5b31ca41",
   "labels/visibility_states.json": "90846339dd3d0e208e4b5c62cfa465df3897361eadecc3b50777f03dcc38f39d",
-  "source/cameras.json": "5ac50d10aec1836dc2aca473d8ea1ea126e002afe994b934e3f06f2e957af7ab",
+  "source/cameras.json": "09c3b0e92cd91fc1836788302554b140c77bd54df3053acddef31005d3172f0c",
   "source/caption.txt": "f8b1481bd751417a286bdf06576b92f22eb6955c477eef19f7efff6f1b6ea9fe",
-  "source/meta.json": "9426351df6f680b5e22d0b18dd25d874cd71033dccf0b89f31929dec89f4cbfc",
-  "source/points_target.npz": "2779eaad4cce555ace6b49ae2b0a5527afc158a604445719b47ff46db5906a39",
-  "source/visibility.json": "5cc135286faddc58028b53b627361d5f1e1d02026b5dfab868e8363042dabc30",
+  "source/meta.json": "5b7d9c8eb2b4bea03357d7411c0075a6841efc17ff4e909eb45bf2c7a1f0ca83",
+  "source/points_target.npz": "57200f3cb72fe37a851a4e83227c9d825f7bb0a2a86a7b023e6c77feb7da3ceb",
+  "source/visibility.json": "298d31589d221cec48cab5bca420e33a74d6d7377702330bbb752a447d7ad478",
   "tracks3d.npz": "e9614f0704b9c92ae4f1a1193c6dacb0421d3825afb6799792c46c42abefe587",
   "videos/view0.mp4": "ea3ca616e508e0f749c09b374f88beaed714e7ec3e3cfa8db2f48566f4b0e7e4",
   "videos/view1.mp4": "6aa9ee37e11af86fa3e0c20a76571c190eebaa9aca729572b0596fcf079274a9",
diff --git a/Simul-Robot/c001_episode_000180/source/cameras.json b/Simul-Robot/c001_episode_000180/source/cameras.json
index db6d94e097fad055e3275f400108d911368da9ff..b80d31f528c2b2772c92ffce2326d4ab29246c66 100644
--- a/Simul-Robot/c001_episode_000180/source/cameras.json
+++ b/Simul-Robot/c001_episode_000180/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_000180/source/meta.json b/Simul-Robot/c001_episode_000180/source/meta.json
index 05625d08a53b43aea91e5a6b6a4a41a9c01acfe4..c9142f9ffd4addec0e2b3f6852818c25961ac4e2 100644
--- a/Simul-Robot/c001_episode_000180/source/meta.json
+++ b/Simul-Robot/c001_episode_000180/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 16]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam02'] 의 비가시 구간 타겟 변위 [None, 2.01] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 16] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam02']: [None, 2.01] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_000180/source/visibility.json b/Simul-Robot/c001_episode_000180/source/visibility.json
index d74e44e05d2ba269bee8921502717ac2a0cb9580..297dfc02c44c4b36a2bfee007a0661aa5aefb29e 100644
--- a/Simul-Robot/c001_episode_000180/source/visibility.json
+++ b/Simul-Robot/c001_episode_000180/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "magenta_plastic_crate": [
@@ -8206,8 +8206,8 @@
       "target": "small_red_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 16]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam02'] 의 비가시 구간 타겟 변위 [None, 2.01] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 16] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam02']: [None, 2.01] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -13655,5 +13655,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_001910/sample.json b/Simul-Robot/c001_episode_001910/sample.json
index 769b8ffbac9f5301340df2131349080dcfc05c6a..f56bc7bc6f7033d10384a86b3bd9e0f3053b0a6a 100644
--- a/Simul-Robot/c001_episode_001910/sample.json
+++ b/Simul-Robot/c001_episode_001910/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "8197247e51a06fe2f8a16c693046357a6a04e84d12c6408a6432490d28f077a5",
   "labels/objects.json": "530643ce61b10c7d2dfdf6861e10c0ff30d1cf168a7424a1ca90642c2c238f83",
   "labels/visibility_states.json": "7a34e1b0e47ac45634dae6286277928d1d68d5a6b3d852b9105f0b3be663f725",
-  "source/cameras.json": "207e6a1407830b6ca596803b8d6f208cdc727bab3168ddadd0e9d811eed5503b",
+  "source/cameras.json": "493a32ac749bd6520e609b99541ca45d3809fbec8d48ed3a231964b15e1da9a4",
   "source/caption.txt": "5e9a6ac953d8621bc24a7f9e2820ae95e4433946466b91f72a3e185adb75178a",
-  "source/meta.json": "76db76e49c8139c734a2b7d39434cd63e9dd91a769cc6d45fb50858e8054d6cd",
-  "source/points_target.npz": "fc1d62f63faa9f6f477f6e0f268ae60d2da6f81b97fef543773b8b91087549ff",
-  "source/visibility.json": "1db1b90328e2ddcad464849a6b9abc471b191f89b298e1de2bc6d29c2020c079",
+  "source/meta.json": "b35b8493dc48424024d87af7546721280aa47c2862a9423f07118576bc55fa81",
+  "source/points_target.npz": "3d4d5f17c528da2b4d63ebd1ab82f01410511b642536bdaba060683ef67db572",
+  "source/visibility.json": "dd53478f32a6a1c47bc5d9408dc113e4b9ffecd4250bb0d8b6dcf74e2aace4b2",
   "tracks3d.npz": "e5c49fdb3c6f73b21054f542774a83094f68dc2535353c21a47630c82d2eedbc",
   "videos/view0.mp4": "bc34432abd55f875edae4766e71be84ee1b2a5f92e59d3cf77594c144b7aa6b9",
   "videos/view1.mp4": "33a5f1505bff11b9317a32e6ac02b623b39e95c002cf823fb49366291a0fe5c3",
diff --git a/Simul-Robot/c001_episode_001910/source/cameras.json b/Simul-Robot/c001_episode_001910/source/cameras.json
index 939f40907153bbb5e43990d0f5b335a3629c2121..dd6d5e143d8238a37aff99af570bfcf32953b533 100644
--- a/Simul-Robot/c001_episode_001910/source/cameras.json
+++ b/Simul-Robot/c001_episode_001910/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_001910/source/meta.json b/Simul-Robot/c001_episode_001910/source/meta.json
index 32e2a70d316c1fd0f98d016512d9af8ea3a6f5ea..dd8d81a3f3006c7e0dbc5985b4dd426b509bdd5c 100644
--- a/Simul-Robot/c001_episode_001910/source/meta.json
+++ b/Simul-Robot/c001_episode_001910/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_001910/source/visibility.json b/Simul-Robot/c001_episode_001910/source/visibility.json
index fda00783e1cd10d5575e3eb4a48ce6cad3af3c2e..6e7c6459739103160f3c9e53c0c28cfe0a9cdc7e 100644
--- a/Simul-Robot/c001_episode_001910/source/visibility.json
+++ b/Simul-Robot/c001_episode_001910/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "cream_ceramic_cat": [
@@ -6130,9 +6130,9 @@
       "target": "cream_ceramic_cat"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -10200,5 +10200,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_001960/sample.json b/Simul-Robot/c001_episode_001960/sample.json
index 5b4fb2fa2802670db1331f837c6ba6e0c21a5c6a..c7c43b081e2903ea4ece1cac4a93d8f1bfeeae7a 100644
--- a/Simul-Robot/c001_episode_001960/sample.json
+++ b/Simul-Robot/c001_episode_001960/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "ad18571db2bc4e84eeefead8223fd61a77268e562e154dcd27053ff74375ce3a",
   "labels/objects.json": "e3d2c66cb518a37216345ee3160f0cda5d10e56b90801e415c57d1121ee9ac3b",
   "labels/visibility_states.json": "9c7ddf78de49309e1ef1adaa13d96bf09ea80c40909f7c35a93307e8945e6b50",
-  "source/cameras.json": "0308df0d22a3349d453776fab0c8feea5d4d77d97de622fb2e89cecc70540cad",
+  "source/cameras.json": "4634e21a5201b6b49b2300d434669b471cb24752d72d33f6756cb63561ed22cc",
   "source/caption.txt": "a2924ffcc09249c4aa1c1512f410e401b9d92e81a54a7d88eacbd06d1932e038",
-  "source/meta.json": "a7c2acc492a93ac90ec8ce7f86ab5465e1cf105f171ae92fe2d3d71b5f60541d",
-  "source/points_target.npz": "1acb4c58b784c5b0ef94a56467cd8925b6484021c099de0375a191fb08b77bee",
-  "source/visibility.json": "7e0ef347848e89b1b1238ab1d2c35346896e6bc746498a488a3c819e5649e389",
+  "source/meta.json": "090d24025d3acc396af7bac2ad3048c2dd37639354870a7572285c4eb53a46e7",
+  "source/points_target.npz": "4f72f33e2d44ede0ecc6ad9d13fa2c1b778b880fc0a0ecf94d2383817737aefc",
+  "source/visibility.json": "f2df949bf101753511c43acb5881c2fec6fff7312e3170eca7a0ba5a9b8572f0",
   "tracks3d.npz": "238a2bea4d0824fd0f8a3bef041ec854f8196ae566a308f5d0a1a882e1626d3c",
   "videos/view0.mp4": "b1c5c9a7311958fcae6b1de917bea5745d655478030d61729ee4e32856a12cf1",
   "videos/view1.mp4": "2efeabf441cb8009f3d4eb605fd07b11bb146db33e60c90f7a3c28a636fd5240",
diff --git a/Simul-Robot/c001_episode_001960/source/cameras.json b/Simul-Robot/c001_episode_001960/source/cameras.json
index 059de885bf3103522fadbf1d59b77f32e1c892b7..f6ed309533d26859a0ef09e0581d3f80f85e2456 100644
--- a/Simul-Robot/c001_episode_001960/source/cameras.json
+++ b/Simul-Robot/c001_episode_001960/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_001960/source/meta.json b/Simul-Robot/c001_episode_001960/source/meta.json
index 470639a1f8b3c641b195b2046378ce5fb58132f5..08e63589975df0f2cf1b439068e444e79bcd4a04 100644
--- a/Simul-Robot/c001_episode_001960/source/meta.json
+++ b/Simul-Robot/c001_episode_001960/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_001960/source/visibility.json b/Simul-Robot/c001_episode_001960/source/visibility.json
index 3da3bd05f88894785418afb9a54de22da38855d1..030425bdb6ccbbf134489c5a436351ffa9b993dd 100644
--- a/Simul-Robot/c001_episode_001960/source/visibility.json
+++ b/Simul-Robot/c001_episode_001960/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "cream_ceramic_cat": [
@@ -5380,9 +5380,9 @@
       "target": "cream_ceramic_cat"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -8950,5 +8950,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_001980/sample.json b/Simul-Robot/c001_episode_001980/sample.json
index 78b0c7009f3bad2f0aa19146d1fc6c230a162289..6dc3dca643ca9ea679a348e9b122046725c0adb6 100644
--- a/Simul-Robot/c001_episode_001980/sample.json
+++ b/Simul-Robot/c001_episode_001980/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "1db691eb71365d819bcf2fc18ff12ae1ed02b9ced7fea5751d7eb316c1ca92e3",
   "labels/objects.json": "34eced3a5df2610c8868cde4f3ec3ddb34aa7f772650a6b9ed537969a4725909",
   "labels/visibility_states.json": "9dc490bc77b060883382dcea75d5c7783daa0617be7e6af90c1704e41953c49e",
-  "source/cameras.json": "b7b15b06bc9d1345174cb6e713fec881ec2ddc66048cd9b30808b08bb6a1c91d",
+  "source/cameras.json": "94b294794f3e1d21ed141de5b940d03025cafd059a034aa9e50f9d3099f64561",
   "source/caption.txt": "9ac05c3bb76b4205b216843200312e16b9f16e64bea9986461e48c7582c07e3c",
-  "source/meta.json": "6b7dd43f24504fa65c114a7a2d3d5aec40c0416385f58728cf5c91505b830a94",
-  "source/points_target.npz": "ad59bf28c8a5b717f9854245f1b1163085900f73129a7f094fa0f3907ca3517c",
-  "source/visibility.json": "772af1295373cc306f018f30ae1f5e20b20f796655ef3caf25d0c537ee8b64ff",
+  "source/meta.json": "ddc856dabb4197424a4107513fc647bbf9a73c7ec94014465959b2a304eb8a4c",
+  "source/points_target.npz": "f1bc41e9bc74edfab4dcaf2ab89cf774d8c28f30f0e685c465adafa12e8aa563",
+  "source/visibility.json": "c85e27548dd5fe167546453060ff7d0eb5c065a439eaf30a71c33350dda01c00",
   "tracks3d.npz": "f907daf79257e88025d0162d4bdc7639a414fe562c22f36047e03bfbc1abdd4e",
   "videos/view0.mp4": "a9f9b82a5c43b57efb8212da8b75da21585df5856a3437e50dc5f4c62d83229e",
   "videos/view1.mp4": "de808c66b38967d2e26a9a19e1451a6d679aa2d870e261da732026f091941604",
diff --git a/Simul-Robot/c001_episode_001980/source/cameras.json b/Simul-Robot/c001_episode_001980/source/cameras.json
index 8e7bc5cdcc9c68fb44c4e39400219cb5025ed3c3..da799bbf5cba0ac8fbc547c9c9d93fa2a22e2685 100644
--- a/Simul-Robot/c001_episode_001980/source/cameras.json
+++ b/Simul-Robot/c001_episode_001980/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_001980/source/meta.json b/Simul-Robot/c001_episode_001980/source/meta.json
index 90c858b02cdb2cf048eefd55437a244c080a0221..8137df0c9c60d0ff525d3e0fb89bd1922ffc50a2 100644
--- a/Simul-Robot/c001_episode_001980/source/meta.json
+++ b/Simul-Robot/c001_episode_001980/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_001980/source/visibility.json b/Simul-Robot/c001_episode_001980/source/visibility.json
index 61c7ebb41c3185e839db8b387a10136fc7aee49c..7e1f72b69b383cb9e95fb39eca3176fbe53471bb 100644
--- a/Simul-Robot/c001_episode_001980/source/visibility.json
+++ b/Simul-Robot/c001_episode_001980/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "cream_ceramic_cat": [
@@ -5590,9 +5590,9 @@
       "target": "cream_ceramic_cat"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9300,5 +9300,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002190/sample.json b/Simul-Robot/c001_episode_002190/sample.json
index 40218c1077217bcb6c195494020d8d918e6a9a52..36ed3a6d80e033f0f31508f72570350923a7b749 100644
--- a/Simul-Robot/c001_episode_002190/sample.json
+++ b/Simul-Robot/c001_episode_002190/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "f768f58537c1816663b8955e6e63dc9faeb993231b425950b3848f0a8ed5cd64",
   "labels/objects.json": "0296b1f062f94c52ae50beccfe5b507e956d0e6c32544b32baae50e1acb7b58a",
   "labels/visibility_states.json": "e72b3cc16ff75d0d8753615536ad74dcced8df23325e4c63a9132e959fe4e16c",
-  "source/cameras.json": "c33321597e87081b95ddcc8324ba0b51c1d91788cb32cd00df7431384a3753c6",
+  "source/cameras.json": "50ee4aa0b204235b18b6939f34b808ce8f957e7911cd36a7237576f35db5cf63",
   "source/caption.txt": "d17866f3c2ad0556b1dea3dd7666010dae72751b2ec1cbc505922ac9d965f14b",
-  "source/meta.json": "6d478d327046b23ffadb2f829d6acb5d53b6b3db379b567c2a061268872684a0",
-  "source/points_target.npz": "1126e12fb72119575798ac21de7c3d1d5c9a84df37606815312c4e5a86c71e78",
-  "source/visibility.json": "673b0c147078d6dfb69d6cb359617c865cc098f5c0886709abca2cce7c857ddf",
+  "source/meta.json": "6e880ad23e625b8f3857193a717eeac561ad49e5578499ecc419f0118357b48c",
+  "source/points_target.npz": "e5449bdfa25da6c6f91a9e563101481ab40747373770af0cdaa363572ba4ea68",
+  "source/visibility.json": "cd779846f5e625062e6cc18a3166af93fe173173b1e98663d59cfc41c6a8c80b",
   "tracks3d.npz": "0f26916716429ab62d41046084eb60e2663b96e4c535f97f0eb25f496368835c",
   "videos/view0.mp4": "df59c30723fedefbbbb1839e91d36769a40ba68e53b907bf20c51531aac9b8ad",
   "videos/view1.mp4": "7fa9fec50766ecc239c6986f027f49594e39537430e4df0b4f70dc63395533c9",
diff --git a/Simul-Robot/c001_episode_002190/source/cameras.json b/Simul-Robot/c001_episode_002190/source/cameras.json
index 0bd27b015341a09b932e668b87596c686f32eab7..68c53e87190b6f4d6b5c2b36781da555fa0a4caa 100644
--- a/Simul-Robot/c001_episode_002190/source/cameras.json
+++ b/Simul-Robot/c001_episode_002190/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002190/source/meta.json b/Simul-Robot/c001_episode_002190/source/meta.json
index 4738471015647f9f38cb930bc50204225a2fce47..a98663896ca5e35e2e4a750e1b9d6722b64989dc 100644
--- a/Simul-Robot/c001_episode_002190/source/meta.json
+++ b/Simul-Robot/c001_episode_002190/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 12]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 0.3] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 13f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 12] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 0.3] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 13 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002190/source/visibility.json b/Simul-Robot/c001_episode_002190/source/visibility.json
index 73781c8d841a07f502811c08b8f4752cc603c62c..e5eb67a1d619930447c7c31398abd598b727a919 100644
--- a/Simul-Robot/c001_episode_002190/source/visibility.json
+++ b/Simul-Robot/c001_episode_002190/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "dark_foam_brick": [
@@ -3103,9 +3103,9 @@
       "target": "dark_foam_brick"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 12]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 0.3] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 13f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 12] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 0.3] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 13 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -5153,5 +5153,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002240/sample.json b/Simul-Robot/c001_episode_002240/sample.json
index 707ca03d928ce2745940c5f9ccdd651d567b1c34..3e09eb485590fe5732c993c04efc84378911838c 100644
--- a/Simul-Robot/c001_episode_002240/sample.json
+++ b/Simul-Robot/c001_episode_002240/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "d1b3201a16eae1076bb50cf859c56ce7b48007ca2b2a79fdb799ab9b8ed2c057",
   "labels/objects.json": "1dd4b90583f93b13e119f3b39af1727598852f544d1e4646de01bf0951ececdb",
   "labels/visibility_states.json": "c9132420419de0126180fd10ae215390467953f0ebc930a80fe6ccc937a06fcb",
-  "source/cameras.json": "3502b668a0268c851ca85cc9714eddad7f8592e7d764c604a563fba731a976ce",
+  "source/cameras.json": "c0067e63f101d6cd34919d7ebe54ab3e42d808c0cb2b4aec1cdac5a15e5ada32",
   "source/caption.txt": "ce61f55bcebc32b15bca8d956e3aa7717450b76447cd05481ebafe1767c147f2",
-  "source/meta.json": "c9aa549f21a3b63fb88dad6af3239524905a86e49165f11d440fa5d176981676",
-  "source/points_target.npz": "2a211a069b76a4a57db3bd9d4b2868b22f31488388bf57e0492d8d43919aaa00",
-  "source/visibility.json": "9ca9ff83fb25834f512e9c985bfd661c09628303e7461cb2e9a1ca1023fa0f4d",
+  "source/meta.json": "8f899f6e942336057f13880e0d201fa075569ebd1c8097d9bff0031bb409507c",
+  "source/points_target.npz": "60f36b5ee7c6d052b101be183c987d88ddd511af20328445004e8f8948a23974",
+  "source/visibility.json": "578c290599152dad692049640a7915387337238ce42915bbbdf069b7b8ea296b",
   "tracks3d.npz": "b1a09fe04af40cce4d721b086ca3010ba75ee1a6d6a785b193b26f239968f5fa",
   "videos/view0.mp4": "2718b4bd3e77893dd283190f5f2553f667b0faa265d464ed96d877b30c1432bd",
   "videos/view1.mp4": "26cd76fa1afca0b75c459d06303649ed0b19ba0608bc1bfc909231f6db616c7b",
diff --git a/Simul-Robot/c001_episode_002240/source/cameras.json b/Simul-Robot/c001_episode_002240/source/cameras.json
index 204bdb6f7635844682cfd8baa014a51c1f7bcef3..248a34f1c0b8225fbfc8922fb498215580d1efba 100644
--- a/Simul-Robot/c001_episode_002240/source/cameras.json
+++ b/Simul-Robot/c001_episode_002240/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002240/source/meta.json b/Simul-Robot/c001_episode_002240/source/meta.json
index 87d0e86eaf42183c3c5e3d8ae34dd95b31d1588f..35e015d04ac3166eeff5fb9617e51289061ae7cf 100644
--- a/Simul-Robot/c001_episode_002240/source/meta.json
+++ b/Simul-Robot/c001_episode_002240/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002240/source/visibility.json b/Simul-Robot/c001_episode_002240/source/visibility.json
index 54230b678e4995455dc69089c459a5b697764e42..f3f560d16fda4de30357f1ba0fae95919ca6c334 100644
--- a/Simul-Robot/c001_episode_002240/source/visibility.json
+++ b/Simul-Robot/c001_episode_002240/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_milk_carton": [
@@ -6190,9 +6190,9 @@
       "target": "red_milk_carton"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -10300,5 +10300,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002250/sample.json b/Simul-Robot/c001_episode_002250/sample.json
index 1556aeb5745cdc58c8888d80fbac7b94f3da17de..ff33f96b0e01ae271a57f08eafed162e5534aa2f 100644
--- a/Simul-Robot/c001_episode_002250/sample.json
+++ b/Simul-Robot/c001_episode_002250/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "ea0d811b34e7b8d286a31d897800960ffee0663e0579af0922acc6daa0e43db6",
   "labels/objects.json": "7e8e4d87f3ddbaa3a0b5d3183f2865b74f9c65f5bfbdb23d825a46272b0912b2",
   "labels/visibility_states.json": "a8072e7bd9e0ef30dccbd1080fce0657a2b1402cb5901d5e691826f8bcc7cd55",
-  "source/cameras.json": "426457342dfeae8c08fc7bc42983692c969f208fb943e1f8e32e6451e7b08cf7",
+  "source/cameras.json": "f2591391052c73037f4bd019f7cf6dc8fb720a4c1ddd192dac1fec666f3316b6",
   "source/caption.txt": "0f95d19299add8a71254aa95b66812681e956c5c2bb9694e664c52e07533192f",
-  "source/meta.json": "63ad28ac7f5cf5eb2a0521e9b6b3739dbd57078250a58c4587c8b65eccdcbbc6",
-  "source/points_target.npz": "5151d420555084af872a9808d33dd2078834638ee878201bcd382f4129b5783b",
-  "source/visibility.json": "2b465bd6ea7a96478faacc5bfd704e1a9c150c070279053dcdecf361c8429c85",
+  "source/meta.json": "cb73e5874fd0b7af719260def45c98ab40068ca3b17fb665eb60cfed1fb39f93",
+  "source/points_target.npz": "e050c859204e0021b6ad3b706c1725ca6bb60608c836010c1e5d2f3aac302870",
+  "source/visibility.json": "db080002df5ae4bae32b3ba02fb3b5bc8f01d06c2d486f643d7df735f256404a",
   "tracks3d.npz": "20051439cc58e209e29996b2807cd1a878a3c86a97e5a54fc78bfc13f95d7121",
   "videos/view0.mp4": "77efce217a7cc56177d0c9d5249eb41bf286972ac2e749ffdce74bd56c555ddd",
   "videos/view1.mp4": "3be1a82d8122078029329610b86294e2ab6a57ede270189801d10c71533cb1e9",
diff --git a/Simul-Robot/c001_episode_002250/source/cameras.json b/Simul-Robot/c001_episode_002250/source/cameras.json
index d93449104cc6d4081c473a76a177c3a3ada6db47..966cae7a005dacffa47606f9a579f251a86afbc2 100644
--- a/Simul-Robot/c001_episode_002250/source/cameras.json
+++ b/Simul-Robot/c001_episode_002250/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002250/source/meta.json b/Simul-Robot/c001_episode_002250/source/meta.json
index a5336c02e253fd11e660c03397d1b7543f1a8dd1..5a59a21312b260f5ba3e3169d425d45dcdedb71a 100644
--- a/Simul-Robot/c001_episode_002250/source/meta.json
+++ b/Simul-Robot/c001_episode_002250/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002250/source/visibility.json b/Simul-Robot/c001_episode_002250/source/visibility.json
index 461859d41c3dd63fa5b2d60a6533ae8eb6edb09e..e51c2ccc1278e3a4ee168dce599dd350db0eca7b 100644
--- a/Simul-Robot/c001_episode_002250/source/visibility.json
+++ b/Simul-Robot/c001_episode_002250/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_milk_carton": [
@@ -2530,9 +2530,9 @@
       "target": "red_milk_carton"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -4200,5 +4200,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002260/sample.json b/Simul-Robot/c001_episode_002260/sample.json
index d164dc9a82a176629d3c314a2410569e7a944298..240c8207677063ce5f4669cd2f076c3142a86d8c 100644
--- a/Simul-Robot/c001_episode_002260/sample.json
+++ b/Simul-Robot/c001_episode_002260/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "2511c3457dc8d5f0bb0fc77b169673d8d08915c5741fe456bbecbeb06b3eca47",
   "labels/objects.json": "3fe9c832b6a6a183dc2398555aa712b5dcd5370e2357bb20e72c23537dcd6275",
   "labels/visibility_states.json": "22f4061ca7248d44ee8252beee8c7d144685524553b293b45d0bf8cabd417601",
-  "source/cameras.json": "e17d7826f3614ac516c9cac0a5d07555e5f935b77356d8539f4546a11fe1af50",
+  "source/cameras.json": "b4a5aa7903a4d32b50a5f767e5eeaba2bc2ef0b54dec847245599d535c9e58a4",
   "source/caption.txt": "028e3f6db7b77b98796e7aee46b116e4d303f4f8240c3932a629c7d7095e87d2",
-  "source/meta.json": "fe42e5852068b8ebbc1d4dd7738538fe78a4f63c207c23ec3cdeb7f21ae7506d",
-  "source/points_target.npz": "69eb5609986d8620ff199d1febc881ac75ac58294aad15e66675803297e8407d",
-  "source/visibility.json": "28a05cdf80fe1eea95357e2dc1952552bfbb1583d7497c01a8c90d22a4aa34a9",
+  "source/meta.json": "0636eced933af49bc764b45b3c96cc3060f79e5068a10b4e6488e4f38ab48185",
+  "source/points_target.npz": "d05818b67883c5bca2425767fb3f843912831641f39b4c33f76b2f426f8780b3",
+  "source/visibility.json": "232d3d345d5825e0de57b73acf6d5b3ab6d5db90285e23ecc11e158b68049844",
   "tracks3d.npz": "bee3b75b25cf5f19837f3d727cb90c4c582879069a7c403bc371c4db9c994abb",
   "videos/view0.mp4": "c9647d8351e4c6c7a3e8184e4276ef3c64f6451d3c3678a3c3fd96af48ec34fa",
   "videos/view1.mp4": "9c11c7f1c763b659e0a6f51cc5b1f3af1e1b2adfc34ca2ac0be7ba8b3971e4da",
diff --git a/Simul-Robot/c001_episode_002260/source/cameras.json b/Simul-Robot/c001_episode_002260/source/cameras.json
index 67f7bbe143744e5fe7f506c52f10ed2c71b27e52..c30b368c18d1596f51fa6c394b8cbac6a088c6cc 100644
--- a/Simul-Robot/c001_episode_002260/source/cameras.json
+++ b/Simul-Robot/c001_episode_002260/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002260/source/meta.json b/Simul-Robot/c001_episode_002260/source/meta.json
index 457d4c9a50b2472554bf33603981ad1fc5b45a63..ae31d617d55b741631f36b5aeab05e7fc0657877 100644
--- a/Simul-Robot/c001_episode_002260/source/meta.json
+++ b/Simul-Robot/c001_episode_002260/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002260/source/visibility.json b/Simul-Robot/c001_episode_002260/source/visibility.json
index 89b60b0277b3107547479a2f62125f14fb3fddc3..78e6918473c832b48e4e17c603bc028e184b3c29 100644
--- a/Simul-Robot/c001_episode_002260/source/visibility.json
+++ b/Simul-Robot/c001_episode_002260/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "ranch_dressing_bottle": [
@@ -7360,9 +7360,9 @@
       "target": "ranch_dressing_bottle"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -12250,5 +12250,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002270/sample.json b/Simul-Robot/c001_episode_002270/sample.json
index 591e848f1c06c04f36c760f935ec180979fe7001..2e5e15472d67a1c99683f5fc8d5a9d1aa9d1b47d 100644
--- a/Simul-Robot/c001_episode_002270/sample.json
+++ b/Simul-Robot/c001_episode_002270/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "7218240cc585bc47ad6b6d4f63d5fa6a814f45d3a118ffb7f2cf9ac9b0b0235e",
   "labels/objects.json": "d507b9174ce154cc379fee2de48d61a67c33c502019e7c5642c1c17e2d9ac8d9",
   "labels/visibility_states.json": "41033b806a390774c0555897178f18cb9ac7e16a755a62de42a133da61d44459",
-  "source/cameras.json": "3d805d895e0334c07d252ea3143527bad89417cc781da720da23b076c1f42c73",
+  "source/cameras.json": "d522700f34060eb147c017b66745d97c20b2e4b2c1ebc0714761334de6d98aec",
   "source/caption.txt": "c25e86073b78d92a15b794b2c023241cfa3c6c9afba63464114a382a4a591fad",
-  "source/meta.json": "11a9e8eb4b6187b08c5a53b4199a0c71867394bd3ef756a75d4d8fba3f1def3f",
-  "source/points_target.npz": "6c4482fd060e815c2ed426dac893af6d2b05f5fa1f30b38830ee394587ffd6f5",
-  "source/visibility.json": "50bc264d1203cafadfa0345fd99ca279aaad0bb673bf47faaf1cbf82aa4885ea",
+  "source/meta.json": "fa0635048b3db2f8ea551b3e1709adbdfccaa2ab6a163b987be34f28dffad97b",
+  "source/points_target.npz": "c8557ee6fa8c188a3b5a9735e687c8586c86ebfb1a3f3e54d3f72ef2bad13225",
+  "source/visibility.json": "5d10982c7527e04baf692d7633e5a4d1cdc8f66607dbd788704865639192dbe2",
   "tracks3d.npz": "37d3de1939e04395a841398c45a77b3c685962ab87413eb7247d9c9bef956b20",
   "videos/view0.mp4": "8b8ce7f7c43b6784b225f7059f1fc6331be4c475e76da65f587681d321ccea4c",
   "videos/view1.mp4": "41e252ad2b28a53536721a1f4aa0c29b2ac5e030a6d00a446af35ab8594e28e9",
diff --git a/Simul-Robot/c001_episode_002270/source/cameras.json b/Simul-Robot/c001_episode_002270/source/cameras.json
index 365e2bf3be3b9a761d70cb3a3c8a1194ab134cfc..98c767beb690610fdf0fc0256a555ff645b17ff1 100644
--- a/Simul-Robot/c001_episode_002270/source/cameras.json
+++ b/Simul-Robot/c001_episode_002270/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002270/source/meta.json b/Simul-Robot/c001_episode_002270/source/meta.json
index 1a92cc372e4d05bc7c78713e511d35f937eb9355..2bcd45d1286f78979b1996684da59b32fb38621e 100644
--- a/Simul-Robot/c001_episode_002270/source/meta.json
+++ b/Simul-Robot/c001_episode_002270/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [8, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [2.95, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 8f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [8, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [2.95, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 8 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002270/source/visibility.json b/Simul-Robot/c001_episode_002270/source/visibility.json
index 60382023ebd827b4e9d080dca756177d27e2feba..d9de1b932a86f23d74fd372c555e2a8918e0d4dc 100644
--- a/Simul-Robot/c001_episode_002270/source/visibility.json
+++ b/Simul-Robot/c001_episode_002270/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_tuna_can": [
@@ -7813,9 +7813,9 @@
       "target": "blue_tuna_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [8, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [2.95, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 8f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [8, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [2.95, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 8 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -13003,5 +13003,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002290/sample.json b/Simul-Robot/c001_episode_002290/sample.json
index c3adf6b64a59673d699f3cc5cd5d1bf37fa01b1c..9fb46195cfa230a311186941bee58921127c4e03 100644
--- a/Simul-Robot/c001_episode_002290/sample.json
+++ b/Simul-Robot/c001_episode_002290/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "4e6fc8df14947505c6603f1947b2884ef8988c2d589f7ed88a2dcc1f5f3232aa",
   "labels/objects.json": "5387fe589149bd807682f7c3c59d8dcaaed127823a8a1e5942dd8bbcb49b083e",
   "labels/visibility_states.json": "3761551e34fe47b69dec9c6dbbc9ff8333984c4592c84e7f040e7a065a7fe6da",
-  "source/cameras.json": "69e3cc47099efdc4706dd0ba4ef52851d61cfbf0ef993b86029e75d173e5fa54",
+  "source/cameras.json": "6a7dd8b30298606063e4d1dfc23fe7714cc2b8ad77741dd6d5d0b87a0a64b507",
   "source/caption.txt": "db51146a6903bc959dc345bbc9b6f87e5f87d2048710ff9b1fbbf092145b8523",
-  "source/meta.json": "26cd3f50c5fb57d1c6cee5cdaa8bb41eb991a45dfb13c955b8426d4d8077d192",
-  "source/points_target.npz": "30a2e7d319ba84164a721be8c7999c5438242bc7d66fd8c884ec0b5b508e1b12",
-  "source/visibility.json": "3c2b282ba9c9b8834d82c8625eded298ca42f63aed5848e943abf6ba79a89af7",
+  "source/meta.json": "5a9db331d3ecdee5d6dd70de8c22f1523a0e410fdf18669384336b75b2b78cfc",
+  "source/points_target.npz": "7da41b7dfe4387d42aefb29417a18ea1faaa86f10eaa700ef070ecec647ff2e7",
+  "source/visibility.json": "6f470ad75e86cef51ebc63528c9974eb51a56eff334f4b3921c20a80cef1d4ee",
   "tracks3d.npz": "f82ac1424725831864d803b2f9b210b00b743f1e986e988f2ca621afbf55fc08",
   "videos/view0.mp4": "3cb68a1a0a810473a227f5f5fbe919eabe68603589d90257464b295b25d687b1",
   "videos/view1.mp4": "f625f80b9d8d9206a66da6f0b3abbe1334ec743dffc6c67eeafbed811e850f6a",
diff --git a/Simul-Robot/c001_episode_002290/source/cameras.json b/Simul-Robot/c001_episode_002290/source/cameras.json
index 0412c8190e8ea764b3af2686287983c1e6e669a1..6cce782c620ed8b86bcdfca9464045a27931642b 100644
--- a/Simul-Robot/c001_episode_002290/source/cameras.json
+++ b/Simul-Robot/c001_episode_002290/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002290/source/meta.json b/Simul-Robot/c001_episode_002290/source/meta.json
index e95035f3eccc7197d6fd71d6a2b411d87f2350a0..5b7fb49af26de1ba0dc62e847ae6153f631e14d9 100644
--- a/Simul-Robot/c001_episode_002290/source/meta.json
+++ b/Simul-Robot/c001_episode_002290/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002290/source/visibility.json b/Simul-Robot/c001_episode_002290/source/visibility.json
index 7f01712c2637d333a087a5bc9cbfdb6427b07ccc..b24cb56c62dba8265f0548ef83d2a26554003aac 100644
--- a/Simul-Robot/c001_episode_002290/source/visibility.json
+++ b/Simul-Robot/c001_episode_002290/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "granulated_sugar_box": [
@@ -6580,9 +6580,9 @@
       "target": "granulated_sugar_box"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -10950,5 +10950,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002350/sample.json b/Simul-Robot/c001_episode_002350/sample.json
index bc5d69ac0ccc90d3b634ed7804e451a5b8a85db5..23dc91aa170f13f59948f1f34b1d99066e025661 100644
--- a/Simul-Robot/c001_episode_002350/sample.json
+++ b/Simul-Robot/c001_episode_002350/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "b44cefdabf8d19fa17b137682131482225af28c706a93ee7281f4b8567cf675f",
   "labels/objects.json": "9202166ea4c692f7acbbf5c167995e957cd42b7af919168dceaa0ac5082cc961",
   "labels/visibility_states.json": "488fcf20ed37e509c14a110634b8efb4b943ee4fba7c7f78a0e9ae9d42c47513",
-  "source/cameras.json": "3205aee91aa6566ea10bd1413381fef09f7f9d6b2ebdd9cdd142c57d1697df38",
+  "source/cameras.json": "22cc71c1f45154c2247ab6926e2222d99fef0f74b1791eacd79497352f6088c4",
   "source/caption.txt": "3bc7f0c72f07b56a87a918dfe6a501b0ed6f3c02ef984d2d2b0c90024995d678",
-  "source/meta.json": "a3fbff23b998493dabd536ca526d1ffcefb84de39a5a7f72131b5bcc9d4ccb0d",
-  "source/points_target.npz": "9d1078521a32b686027ae6481c6b3b8d254741c101d5fc423cf4509784a4f8cc",
-  "source/visibility.json": "fc3125953f7b99362a1ac232170646cc698dfd7b27c57defb912dbd49db3e9fa",
+  "source/meta.json": "576f26e3ee384cefa62320e42a2151a500f3d7328f88c47a9af9f359e5967cda",
+  "source/points_target.npz": "281fd07cd76e314b174c5a215c38e5fd977dabd5f7824adc57520be777704319",
+  "source/visibility.json": "db7de1c16016288aea6d98cde7d4a8b80082d050d1c6a2b2d9cf14e19f30c1e6",
   "tracks3d.npz": "8840d08d4cf7a8b24c29ab0ad86a5a3e40d868706cd50781c9d7927038dc9a1e",
   "videos/view0.mp4": "288ca08b93a318b6cbac641e8cbe63fb8784fc4e2ad3eec43d2715c33ee89b3d",
   "videos/view1.mp4": "43e57e67cc38957d08541bd5e149533a5d6b1ed5d2bf008fa512b2715e877c4f",
diff --git a/Simul-Robot/c001_episode_002350/source/cameras.json b/Simul-Robot/c001_episode_002350/source/cameras.json
index 63e6d3bac35c3bff195f3e755d7c76169604d2d3..8c66b972cd0088835f98c7e57f180a66b5c8b61c 100644
--- a/Simul-Robot/c001_episode_002350/source/cameras.json
+++ b/Simul-Robot/c001_episode_002350/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002350/source/meta.json b/Simul-Robot/c001_episode_002350/source/meta.json
index eb55281881acda2148f939a341e1c2c7c857bea3..f664e77120796f03a000e7027deabfa81457e702 100644
--- a/Simul-Robot/c001_episode_002350/source/meta.json
+++ b/Simul-Robot/c001_episode_002350/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 2, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, 0.82, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 2f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 2, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, 0.82, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 2 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002350/source/visibility.json b/Simul-Robot/c001_episode_002350/source/visibility.json
index 1b69d7d969dfc578683cf5feb81ea91bf5010771..db992dbbb82231ce0854a368880130f0bdd42701 100644
--- a/Simul-Robot/c001_episode_002350/source/visibility.json
+++ b/Simul-Robot/c001_episode_002350/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "bbq_sauce_bottle": [
@@ -7243,9 +7243,9 @@
       "target": "bbq_sauce_bottle"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 2, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, 0.82, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 2f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 2, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, 0.82, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 2 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -12053,5 +12053,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002480/sample.json b/Simul-Robot/c001_episode_002480/sample.json
index e1d7072c95af96f62467def8272355cad7356683..4623cb60f17cd297b65bf5bc4d33f17c87257aaa 100644
--- a/Simul-Robot/c001_episode_002480/sample.json
+++ b/Simul-Robot/c001_episode_002480/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "946715cb44b26b420c8f408aaf0dddea79b09424d5a36c9b3bc61168121e5232",
   "labels/objects.json": "0cc0710f4b30e53ebf0f2b0598f83cbd7ffda0e0f3bbf8dc1e9b58a14949f379",
   "labels/visibility_states.json": "cdf4c3bbeadf57777728b396a11ed0c7228db31178795fdabe784e1daddfb725",
-  "source/cameras.json": "0a97f52c8fe01d6c85583646e5b7c92fadcd7b8977b2f11602b735451135f676",
+  "source/cameras.json": "2a3df87fc9933d9d6266cbe40ac773732187d0b27bf8132a4f4cf1d8ba3f9502",
   "source/caption.txt": "3eabeb3d43b0f4cb209cd1db1b35896820e26e8b5c09b9741d9a8d19703d325a",
-  "source/meta.json": "5512aef56b00dca31a8813a2643e7c9f6bf74c8e69c6ef4bb67f23ba1632bb54",
-  "source/points_target.npz": "879397f13f266cb0aad5d25c49e3c6d5fccdfdea0634a2a0d469d608589ac114",
-  "source/visibility.json": "2b4707db5c5e979a872339dd0b3c4248efb852c67fee6a6a63266d30517aa888",
+  "source/meta.json": "00ccb4fa9945d8bdd4f27204cb55e0e93050d9d6ba8df1ddf6e9da1b73d73b15",
+  "source/points_target.npz": "5d0aa21fd489fbc1c5313e3e3004ab4ed0c13f3090a0b6d823c908d4d881bd30",
+  "source/visibility.json": "7acca293fd28deadb39e0ff8c0833b5fa9eb0ad10be30b9894d093b05258cfcd",
   "tracks3d.npz": "3e39bd76790f410a707418af53b6a520caab84b5783080e0ae9b88646193ed0b",
   "videos/view0.mp4": "1d3c5a85248c37a550da24795e9c31f334e487ba18a4713b734afa6d919b43ba",
   "videos/view1.mp4": "04e153e734eb58d40781e486a69fa3df9f1c7173dcc510e6ce2d5f31a8738a3f",
diff --git a/Simul-Robot/c001_episode_002480/source/cameras.json b/Simul-Robot/c001_episode_002480/source/cameras.json
index 53c0fbc7d75334a32a0df18df60594452756dffa..73b1ee65b8cc749c7ace576abb7023a47577fea5 100644
--- a/Simul-Robot/c001_episode_002480/source/cameras.json
+++ b/Simul-Robot/c001_episode_002480/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002480/source/meta.json b/Simul-Robot/c001_episode_002480/source/meta.json
index 109c90cf80e6bcb95dc944260fd2a25cd0b47814..47b85c1cf844041b578c251541be89a195657fc8 100644
--- a/Simul-Robot/c001_episode_002480/source/meta.json
+++ b/Simul-Robot/c001_episode_002480/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 5]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 0.0] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 10f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 5] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 0.0] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 10 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002480/source/visibility.json b/Simul-Robot/c001_episode_002480/source/visibility.json
index 45b66d427eeca77cb305d0371397311adfd55302..abe8113a240466ae4ad720e36b903887216c4d43 100644
--- a/Simul-Robot/c001_episode_002480/source/visibility.json
+++ b/Simul-Robot/c001_episode_002480/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_tuna_can": [
@@ -5533,9 +5533,9 @@
       "target": "blue_tuna_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 5]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 0.0] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 10f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 5] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 0.0] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 10 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9203,5 +9203,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002640/sample.json b/Simul-Robot/c001_episode_002640/sample.json
index 7eb1b99d648a91de40c5f0d159997dbf8fdd48db..128f3a579c6c2be2bab8346abf5843fc389b1b67 100644
--- a/Simul-Robot/c001_episode_002640/sample.json
+++ b/Simul-Robot/c001_episode_002640/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "dd9939383a1ab4acaca1ed2473459a4c250894e8cec4733705005a9575a2e4de",
   "labels/objects.json": "459b9fd8c3d83faaa68c8c0b5f3ba6c3c8e6554c9ce31e39e83d8c036e83bc92",
   "labels/visibility_states.json": "0e77a57d19ac60bf60d29fd1a30f38cce29d91520f5205748754ef83cb5cb1c1",
-  "source/cameras.json": "b3b06c1663ce71e4ef6fb20fe8a86dc7502db8d7eebbe8ad6bbc536817b491d5",
+  "source/cameras.json": "7e4b744990dda302ca09d0889e8e1bc8ea55702c041d8f9a48e14d8aee23401f",
   "source/caption.txt": "ba7d35e3d3feccd0c90571eeb1103c9c01d13f1dcd4e2a0b40cb7a568055a82f",
-  "source/meta.json": "db73407cae1806c43cf41c5731665947d5d7884dfdb4d54612408f7fbfcbd1c1",
-  "source/points_target.npz": "1a7e10d9e640e33e0e7c51b2177e91145b4dc9aa2d84127d46c3bf7aac978655",
-  "source/visibility.json": "e51ea39235ce4f8e9f97111003845cc68615b01ba02d37730e26a4bcaa8fd9f5",
+  "source/meta.json": "f8bc875b7aff9d90c555d497ed693c8d8741a876388bf73d99fdda057725382c",
+  "source/points_target.npz": "228c1266b648662903d9df32d7271a8e7380e9f2d2a649abdfe37f2b7db8c3e7",
+  "source/visibility.json": "85776ffbcba4b687aef10b771847e530478fa1a394b6e8611ee562bdab952933",
   "tracks3d.npz": "260c08dd4a6b4ee78ef6ce0d18e4e0f67159c7dc8dca422399d2aa391a82acfc",
   "videos/view0.mp4": "5ffbd4737312aa90b1483003c954c0b9cc60e389c614991d2d933239857d4e99",
   "videos/view1.mp4": "2f4e8dceafc8f96529fcce7d1f36457a11493a73083419e9014622ffd440f97c",
diff --git a/Simul-Robot/c001_episode_002640/source/cameras.json b/Simul-Robot/c001_episode_002640/source/cameras.json
index 69423178a50c3afe1557c602d19a717bd0496023..dd9658bc7291f97badd1e42661db005b613ddf6b 100644
--- a/Simul-Robot/c001_episode_002640/source/cameras.json
+++ b/Simul-Robot/c001_episode_002640/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002640/source/meta.json b/Simul-Robot/c001_episode_002640/source/meta.json
index 802a6e53aef53a29458c131439a47ad1b3f0b5cf..2bd3bf503c42bbacf4b5af3619366563063b53dc 100644
--- a/Simul-Robot/c001_episode_002640/source/meta.json
+++ b/Simul-Robot/c001_episode_002640/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002640/source/visibility.json b/Simul-Robot/c001_episode_002640/source/visibility.json
index 82f663648b0b8b1e1cf10933c298b142a14cf265..d37dbe16711048574807d41e15bb5d4ece19a9fe 100644
--- a/Simul-Robot/c001_episode_002640/source/visibility.json
+++ b/Simul-Robot/c001_episode_002640/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_frog_can": [
@@ -4990,9 +4990,9 @@
       "target": "blue_frog_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -8300,5 +8300,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002690/sample.json b/Simul-Robot/c001_episode_002690/sample.json
index ccb40c337100398be4e719720347071208a0056f..720473718de3dd48579b09c82ded366b97e42550 100644
--- a/Simul-Robot/c001_episode_002690/sample.json
+++ b/Simul-Robot/c001_episode_002690/sample.json
@@ -222,11 +222,11 @@
   "depth/depth.zip": "603cce63e69042149f756b3cc2cabff5b71c822ff250cd2997e73505cca94b4e",
   "labels/objects.json": "69b5e39625e37829f63fd59e21d8d0ca1120bf44c28e16ae0f02d51ea947014b",
   "labels/visibility_states.json": "317dfad831ca8e0f020fbddf901c9e0849c40e9bb025e8bf86ce3a49b7d1cbb7",
-  "source/cameras.json": "1df518bfcca204c81aab16a7da87492f172103ce71004d3b2d2c829c591f830d",
+  "source/cameras.json": "25f859c666e962f4d80af85d135088dfd1135d49e10b1a9cabaf38ade8c48945",
   "source/caption.txt": "21a45f87775a95552d1f8bdd51d2594b05248ae21f4de08afb16594dfe101f34",
   "source/meta.json": "60a61c22123d4781dd71875b71412054136a08a093cd4bf32da991f9510f62b0",
-  "source/points_target.npz": "d0ec8ad06d65cc618fac03009654103b880397e739e1ba0634e14fe8cd36fba5",
-  "source/visibility.json": "bd89618f7035d3c8eb61a2d66feddaab610f9d4bf3e8416f4449f2c009198584",
+  "source/points_target.npz": "bd4a7fb221cc0d81da5bba0dd0f02fb2845f38099c3dbef657911bc86257671c",
+  "source/visibility.json": "1f49e950517990a808bb4fee290b7419dc532e8875e1284f3fbe125229ff15d3",
   "tracks3d.npz": "31aecb4f78bdde3f608803e9d590773b05f49d9c7d866f48d0d7203bb422b28b",
   "videos/view0.mp4": "ccdd8125a570de39ee7075bd8db415272bfdff0ad1c7d0a29e7bbd23871d7333",
   "videos/view1.mp4": "0ce64e72c49fac7f33a6e0319f4dc177994e9623b90171ba5b04fa36f2f05412",
diff --git a/Simul-Robot/c001_episode_002690/source/cameras.json b/Simul-Robot/c001_episode_002690/source/cameras.json
index 865b9c529363e9268d9ecac5db9eb65ac474af55..cca433b4e52804b64f127eacb6c492f4fb635dce 100644
--- a/Simul-Robot/c001_episode_002690/source/cameras.json
+++ b/Simul-Robot/c001_episode_002690/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002690/source/visibility.json b/Simul-Robot/c001_episode_002690/source/visibility.json
index 42b72fd86baa71712db7eccd208d4e6ca8e161cd..5eea33d83154708be36110e34b2d0f33c45f47dc 100644
--- a/Simul-Robot/c001_episode_002690/source/visibility.json
+++ b/Simul-Robot/c001_episode_002690/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "magenta_plastic_crate": [
@@ -11695,5 +11695,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002800/sample.json b/Simul-Robot/c001_episode_002800/sample.json
index 0c170dd54893e1f27d06f9a1a175ee260a83b263..bae892dc3ff3ab9aacd645fb19f6f902a65535a1 100644
--- a/Simul-Robot/c001_episode_002800/sample.json
+++ b/Simul-Robot/c001_episode_002800/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "191fa62695370bbf1ef99f57c1ed31b713ba4b502e0bb3276958d838c2e5011d",
   "labels/objects.json": "c6e244d3309b65a032638102bc64a2f48f6727cfc390c597c45fa086d9ce87cb",
   "labels/visibility_states.json": "09c878bc458649a52373dde2f5c5e424c42f1bfd6c4a636da6dc079d0caba8c6",
-  "source/cameras.json": "77574254d5382842c4aa4e42bc12b48a805e9255b00bd648ecc25ddb1d6c05e4",
+  "source/cameras.json": "792dbe1cc825d7796179a96dbd01715c1e94a173fa3859a4669889cb41cbf1c8",
   "source/caption.txt": "611f01e8a4a90af77f8570b5220c6dc2a1ec404fd63aea7920a5611ff779e7ef",
-  "source/meta.json": "7da3a34b518ff139245ad6ccbf2afe572027f8286e320f67d76c022e5a4707a4",
-  "source/points_target.npz": "827f63ef336b446e47b5c8e929907d106e3bbd945719a7361f86bd160ed85658",
-  "source/visibility.json": "3ebfe8250c2da2d8a25e100c5fd6c99e0654cb831bf1e80be022f099dbf2f603",
+  "source/meta.json": "3da93a7ff5e203bfd8d6999a8b66194bbaa3abc456d450d0f08c0f3a6135d561",
+  "source/points_target.npz": "04905d94d4809c0346fbc7d20ffcce7f9f7d80c3508c0d77edb8856d69699f17",
+  "source/visibility.json": "b37a76b02b0b4460c453c9d92f8e9f280bb008a850bdcd72d8d51287db98c940",
   "tracks3d.npz": "90e2d6b9057e57c2c3e8ebf3b96a41026fa51d1fb012a35cc44f474db338dffb",
   "videos/view0.mp4": "a9f632a95352273a00be312259a91bfaf8bc5fecc0c05b527db7d8f40ed06137",
   "videos/view1.mp4": "76190bf26d92dbb731cfa14aeba5c8d2538c5a2d69eebab3b5960702f07a4c93",
diff --git a/Simul-Robot/c001_episode_002800/source/cameras.json b/Simul-Robot/c001_episode_002800/source/cameras.json
index 87da1ea15d69ebb7cecdc213648c78ff1e7b4e75..de26efe9f6a6bd011f65b1634c15ad8135387bd1 100644
--- a/Simul-Robot/c001_episode_002800/source/cameras.json
+++ b/Simul-Robot/c001_episode_002800/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002800/source/meta.json b/Simul-Robot/c001_episode_002800/source/meta.json
index ee6f892400df8bc2cca0a5f02d5187c21002cdf0..5129663808c68bbd834d81a579eb93e5e061b4cf 100644
--- a/Simul-Robot/c001_episode_002800/source/meta.json
+++ b/Simul-Robot/c001_episode_002800/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 0.78] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 0.78] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002800/source/visibility.json b/Simul-Robot/c001_episode_002800/source/visibility.json
index be30bd1796d428dbcd6994d3427ac7779c6ecab5..f2b3109ccaa06a8e8465cc73700efd44d6b0c45e 100644
--- a/Simul-Robot/c001_episode_002800/source/visibility.json
+++ b/Simul-Robot/c001_episode_002800/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "magenta_plastic_crate": [
@@ -6163,8 +6163,8 @@
       "target": "raisin_snack_box"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 0.78] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 0.78] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -10252,5 +10252,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002810/sample.json b/Simul-Robot/c001_episode_002810/sample.json
index d24345ed3591c1252e0baa70d18c227a56014879..d028d4470fca2d5f374ece8ecfaa7ffe6c234d5d 100644
--- a/Simul-Robot/c001_episode_002810/sample.json
+++ b/Simul-Robot/c001_episode_002810/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "a69242f3d62ffe576d6cd85f04925ecee054f0d43e4c657b2f39249a8ef489c4",
   "labels/objects.json": "5a188bfd8e10e6fa9f55741297421b67f59f13364cfa3753d8f25b31822c5f4f",
   "labels/visibility_states.json": "74ce7b930a7717002b58f29f5cd117443e223959247284ec1b5ecc24eeadf9dd",
-  "source/cameras.json": "6afa1223a87cd4e4f193df5362d2e02210662f67d40fd1cefb67089757ae708e",
+  "source/cameras.json": "1c04e583878a181f5d59c8539d4b25c278ab72883f0e8aa91d921f207caa7bcf",
   "source/caption.txt": "b9581384b74f93ea8f32d48e01db7d35e74d7833b9e8804c76622e2f0f032d77",
-  "source/meta.json": "23aecfe9be34d9d125deb8510f455107fd5400c78e33bb4f546283125f099dd6",
-  "source/points_target.npz": "885b74022c85c017df085343ad5f62e6520e28d7e81f4125c4446406ae58dac3",
-  "source/visibility.json": "a170f86aaf1f360ecf4e6e28be8358e9b537758b703edcde82267061c52627a4",
+  "source/meta.json": "ee5f80a41b255e726525e1dc5f0ff76498dfddfffad308af6567e9116b2c3d5d",
+  "source/points_target.npz": "31a9dca016def5f23de8351672e1794a402ee8c148c7ce1d819f30d912c7672b",
+  "source/visibility.json": "bbf5e37b3573696a14b7d7850eb7e1e0607d384efecc6497c1fe244334589cd7",
   "tracks3d.npz": "e0497236b5596dc03ebb11f689b14f475d9ec6d78bd6419f73f9d5ccfcdad5a0",
   "videos/view0.mp4": "0c12f2ad49454701fe49784a9fe2c84921dc5c36f4483e445f6c6480b45e4262",
   "videos/view1.mp4": "40d8b43597f2cf5a8d2388be60ad5ce79c1501fc0a8274b2abc603e52e8fce50",
diff --git a/Simul-Robot/c001_episode_002810/source/cameras.json b/Simul-Robot/c001_episode_002810/source/cameras.json
index ffe41e417d424ae7bc076e0d3a42c980f40a126e..92e8804ca209340d6d777f4772f5225b4eb34d12 100644
--- a/Simul-Robot/c001_episode_002810/source/cameras.json
+++ b/Simul-Robot/c001_episode_002810/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002810/source/meta.json b/Simul-Robot/c001_episode_002810/source/meta.json
index 61deaabb149cb0a38f1c9281e68c49a06549ae3d..cc5ebcd88cec135d4770f9c198e623276f83ec68 100644
--- a/Simul-Robot/c001_episode_002810/source/meta.json
+++ b/Simul-Robot/c001_episode_002810/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [2, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.02, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 4f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [2, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.02, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 4 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002810/source/visibility.json b/Simul-Robot/c001_episode_002810/source/visibility.json
index 8a59d2730038000c47564f66dc8181e45e8887c5..b09399637807612e5b911573c4451a7e8b0059f9 100644
--- a/Simul-Robot/c001_episode_002810/source/visibility.json
+++ b/Simul-Robot/c001_episode_002810/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "black_energy_can": [
@@ -6793,9 +6793,9 @@
       "target": "black_energy_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [2, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.02, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 4f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [2, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.02, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 4 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -11303,5 +11303,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002830/sample.json b/Simul-Robot/c001_episode_002830/sample.json
index 8166f02bdba2e997c9708b81b9d832ab1636145f..330d94dd253e1fdc1f342bc8cd35193d34527ca4 100644
--- a/Simul-Robot/c001_episode_002830/sample.json
+++ b/Simul-Robot/c001_episode_002830/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "268ad3c6dc25c9f6a2ff17fdfdf65509376347e5c9c7f27bdadc33616822a69a",
   "labels/objects.json": "4f6b39037a5fd5129194622e82e4678f48aaa9bd58421113ef3cae9bd95aa88a",
   "labels/visibility_states.json": "49501f7c088071831215030f6d439b2889c9ca9dbb5eb1ec37b45a8db939fb04",
-  "source/cameras.json": "2d465f79d24efe93fb5c76a7cf086f3911e67ec90f1ac998de28b123b5911dae",
+  "source/cameras.json": "c60543e763fb22b5d90ddd8984ee8d18736e8d15d3f8319359994d11ba4aa04e",
   "source/caption.txt": "be149af074fe95ba1abbd810dcf63e95e275e1930807b09a31fbb6f110855c1e",
-  "source/meta.json": "961ca5f5518bffb4070e411c521325fa330e726402c385d742a727368b8f8222",
-  "source/points_target.npz": "1d6c3079bb93dc5cdda7d71902cf933d51ccec75d0fa7910101a72677a6e0166",
-  "source/visibility.json": "8eb5642ca2a27a44011d3d4bd4de491d75976d186b5133ab712f4abfdc9b686d",
+  "source/meta.json": "0431de6ddcc8c21af7c851adbcbe4851d8dd8a49990e2a0f31dacf7f50bcea81",
+  "source/points_target.npz": "a8de617ca5684af6fb3b07c9c711127ce9b028c0c7d680593c4cdb51463dd1b7",
+  "source/visibility.json": "350f2e9c298d049ad862e39aba818cc8f16918b20bffc41324c8fd52997dce35",
   "tracks3d.npz": "6d70ff28174137d7da05e7eb28912d22ce1489e8c2d60bf6c7663c5eac7fbc0e",
   "videos/view0.mp4": "6f9f4d52b8bbdf12d46a6c4e5e8dcfbe3d2b1e59605c74ba801052b8335115a5",
   "videos/view1.mp4": "68d23f5df4d97fb6dcb332461356e6dbb0fea9da144b63c0840f5fee483a147a",
diff --git a/Simul-Robot/c001_episode_002830/source/cameras.json b/Simul-Robot/c001_episode_002830/source/cameras.json
index c5d8abe37dc45a0ea82fa5d9efca7f0c2a18e96a..3dd2ebc56db27dd50aee04f90169ee7e0fbe9fa1 100644
--- a/Simul-Robot/c001_episode_002830/source/cameras.json
+++ b/Simul-Robot/c001_episode_002830/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002830/source/meta.json b/Simul-Robot/c001_episode_002830/source/meta.json
index 865aea3ecbabc0101ca8e15ff0f1b0438406f38d..3cbebe29485c0d41eda8e657e0d71681b6ed09fa 100644
--- a/Simul-Robot/c001_episode_002830/source/meta.json
+++ b/Simul-Robot/c001_episode_002830/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [2, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.29, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 6f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [2, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.29, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 6 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002830/source/visibility.json b/Simul-Robot/c001_episode_002830/source/visibility.json
index d3d90cbb258614817ffa96770fa4034b56c79bbd..24bbdcfb643bc817e0a9e36d2452cc1c96203f0a 100644
--- a/Simul-Robot/c001_episode_002830/source/visibility.json
+++ b/Simul-Robot/c001_episode_002830/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_plastic_tote": [
@@ -6493,9 +6493,9 @@
       "target": "tomato_soup_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [2, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.29, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 6f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [2, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.29, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 6 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -10803,5 +10803,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002850/sample.json b/Simul-Robot/c001_episode_002850/sample.json
index 9793b7aa49d5d976daa8a2ee355a4dfd48572880..152f04cfd2ef842a410c98879ec10a1efa59777a 100644
--- a/Simul-Robot/c001_episode_002850/sample.json
+++ b/Simul-Robot/c001_episode_002850/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "7df213a58dd691a85007930cd1ad42285ed483111c6e3f0af01a59b023f9dad2",
   "labels/objects.json": "30756775255a70dde578c8912082d4d590b328881c26b658dd9a36dc5fd9e442",
   "labels/visibility_states.json": "7b8f003ecc0d5a6081cd1e3a96ddb58b6c3a4b6c1d3ddcf84447b1ca69c06583",
-  "source/cameras.json": "b85ecadbf454f1945c83f1c1e4f604296ea9d02933daa0787f05f3feb100774a",
+  "source/cameras.json": "45968279e0c028b6a5c6a619a63c04498d133676a68f5076b8d39537ee239330",
   "source/caption.txt": "f86663d51445a647d620d7696df0de6c3734208c92086341997ffaa7fc86e6c6",
-  "source/meta.json": "2e0e24914c6547289141db9e6fb9edc99dce457c058eba4bcaaafa16bca40c01",
-  "source/points_target.npz": "f1b6b0389ad9b8817c1ab0d4cfb97bce6b97b0185482724b0ddc716ddb1861dd",
-  "source/visibility.json": "ccba9ffaba37f6a73b9661c7afc24d6dd1dc5f94bebdb0b08fc95cb58c3d2739",
+  "source/meta.json": "a02072a917be646579a72133ceedda81b85e42768a19ef9183dd17cdff828073",
+  "source/points_target.npz": "6c5eb07c896e117634b4856409a8fd5b3468fd3ed0237e918d367837ca0872be",
+  "source/visibility.json": "842465e323ee470dc4b369b1480d8dbcc3f35565ee73e5992efe34960500f826",
   "tracks3d.npz": "f79c487ffb6cd67ce52d77a46a512e84fb0774bcaa7f898567f7e71b28bdab86",
   "videos/view0.mp4": "53a833dc7dc8094f62958f824d978c079d07e639b735b80c413d2985f1dc06de",
   "videos/view1.mp4": "7f24b32338bb4a1db293eaab317d492debaeea51e63bcabe8b4adeda16132ef0",
diff --git a/Simul-Robot/c001_episode_002850/source/cameras.json b/Simul-Robot/c001_episode_002850/source/cameras.json
index 7a7cb9158c41fe2e4f472be7430d0fcfcb486cd3..803e4cbcbdfff8291b72246ffd0b953207f68e54 100644
--- a/Simul-Robot/c001_episode_002850/source/cameras.json
+++ b/Simul-Robot/c001_episode_002850/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002850/source/meta.json b/Simul-Robot/c001_episode_002850/source/meta.json
index 6529729e1ccf7379988d95fea205d5937521b0a5..f25e1d56d85b4a2d85ae089a0e6e00062a12fcd5 100644
--- a/Simul-Robot/c001_episode_002850/source/meta.json
+++ b/Simul-Robot/c001_episode_002850/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 0.98] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 0.98] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002850/source/visibility.json b/Simul-Robot/c001_episode_002850/source/visibility.json
index 00ce9ae1d35a16219bd89316ad6c55077232a1eb..05e8c3c5d0cd158f310c71ee34b4a0c5c3f5c0db 100644
--- a/Simul-Robot/c001_episode_002850/source/visibility.json
+++ b/Simul-Robot/c001_episode_002850/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "tomato_soup_can": [
@@ -5593,8 +5593,8 @@
       "target": "tomato_soup_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 0.98] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 0.98] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9302,5 +9302,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002870/sample.json b/Simul-Robot/c001_episode_002870/sample.json
index 46d1dc435596d20313de8090a7e7f9acab2e23b3..df581bd3921d32a7497f9144db5588e52f342213 100644
--- a/Simul-Robot/c001_episode_002870/sample.json
+++ b/Simul-Robot/c001_episode_002870/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "e181feec8983db99a1f58686f889c32296c84a987675524828e2145e29be949c",
   "labels/objects.json": "97407cf7c43a3b0272962a411f63d8b997055e895ae0b602540c81bb19258596",
   "labels/visibility_states.json": "93ddff875a3489106772d87df73c817d0144a3b3689fbcbaedfb34be91e7d4a0",
-  "source/cameras.json": "d2d8a8c790f14a30f06d1efeb4b4dac637509f28f018ccf78fbb4a43e13322b8",
+  "source/cameras.json": "40187e2277ec5679eaaed1b6b4cf35c65e211c05fa7fb6ee968942c5f23b8537",
   "source/caption.txt": "63544bc72dc2155ebf623182dd3470234098a5d8fa9df4b356f357f7072db07a",
-  "source/meta.json": "3843b1c4ee3ccf3637855defcfda1dba93374303fc306357c4567f97f31169eb",
-  "source/points_target.npz": "bc5f54892878ad63f20dd32af722235fb07e449b71571c034433bb93a14f464b",
-  "source/visibility.json": "d0122a81222ccec0e63890cf5b1155aacbea6703060992d4fbbf62ae0ff91714",
+  "source/meta.json": "2a7f311599140e00425df4205080207c870e3e180315807e1eb9b8979e2e4f8f",
+  "source/points_target.npz": "8edb9ee0a6b0c2b442afe15f010aab0ad367904d09024879e4122db016033400",
+  "source/visibility.json": "41a3cecb42a676b0c5b7a8e3e75a7fc5d681b0ff69bc21b2ea3f4d97b6b3b9be",
   "tracks3d.npz": "15997596e848a53c37940214274dbf3dc8964aae3036ae2d0011969d46576c85",
   "videos/view0.mp4": "7f01477e2e82d8353ff6cc2a109d4a40856f9240d5c3e80d600fc38588149599",
   "videos/view1.mp4": "c3dba6d9a50c2d4339a1cd10737cdc6639e76279e5426249b1c27b9ad99a9532",
diff --git a/Simul-Robot/c001_episode_002870/source/cameras.json b/Simul-Robot/c001_episode_002870/source/cameras.json
index 63b65abdd4c8b6be31a9a08fb43ae03509b13c90..029567029a17bb20282e1de42185dd197fb14531 100644
--- a/Simul-Robot/c001_episode_002870/source/cameras.json
+++ b/Simul-Robot/c001_episode_002870/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002870/source/meta.json b/Simul-Robot/c001_episode_002870/source/meta.json
index faa4faef7c20bc0e452642fc4a01d646d12d30fc..138582c0ff5a36fc6848ad17e90c5149a891aff8 100644
--- a/Simul-Robot/c001_episode_002870/source/meta.json
+++ b/Simul-Robot/c001_episode_002870/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [11, 0, 1]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 12f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [11, 0, 1] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 12 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002870/source/visibility.json b/Simul-Robot/c001_episode_002870/source/visibility.json
index 6390e1a5cf796d7f60df8500235ed5f53b57fad1..7447562606c284408259b2f75953c60d853a7b08 100644
--- a/Simul-Robot/c001_episode_002870/source/visibility.json
+++ b/Simul-Robot/c001_episode_002870/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "dark_foam_brick": [
@@ -9256,9 +9256,9 @@
       "target": "dark_foam_brick"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [11, 0, 1]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 12f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [11, 0, 1] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 12 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -15406,5 +15406,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002900/sample.json b/Simul-Robot/c001_episode_002900/sample.json
index ec870b1f0de050e32d4df8af7ae24c45310ea132..4b1cddfa20479455ce4a9609ca6ea287f0fa7ab8 100644
--- a/Simul-Robot/c001_episode_002900/sample.json
+++ b/Simul-Robot/c001_episode_002900/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "4cd049b71cd6b529347677b3ac25cd93dab8dbfbdcaaf2164cf17b359737a901",
   "labels/objects.json": "7bb56230da12a07e072e9fe4f5714cf2e1e0df44aa4aa3bc7af190beb85d9331",
   "labels/visibility_states.json": "19a38078dcfda8b0d3cd0baad6f0e93b051106e96c8dc9dced9a1d3949edd0f6",
-  "source/cameras.json": "f75198d049e74c67edf26b0d0f4875ce1f35019e07aa1dadef91b68565ad17c0",
+  "source/cameras.json": "145c5e58213a50e6779c1e8f35b719a7a1d0d464047113a845d3fbb291ca33a7",
   "source/caption.txt": "cbbf2f3cfb6b5ce72186fb9f9a201dbbaeab81903a0833139a8526e067f91362",
-  "source/meta.json": "e4348f23581e4cdc64583560caed0271b49384ac4a15bcf956a7e0999ac0bfbe",
-  "source/points_target.npz": "7b85885b3ac16e4fcbda47ccf07cf4ddfea6a38d11b3e99faecb425eb9317b6d",
-  "source/visibility.json": "845d5eea6919cb3539f991ba34c60e193588292c998aa9bf00b439edca25f447",
+  "source/meta.json": "85854fd44ceecd7972d0e577bd9c79f072e4a2b0d838316c1623f25375f1c101",
+  "source/points_target.npz": "5f75d9d78980dafe9a05a62ea4e96eb2d661b26286434fae4d28fb0f7d679487",
+  "source/visibility.json": "0e421e8be0e1142df6fcf17614db69ce9b0efe1b5739038350947bcf71c6028f",
   "tracks3d.npz": "c75eb95dfb8841f073c11cdf52fa3e8090f9b0abecb48a40fdca6e3189edba24",
   "videos/view0.mp4": "1e5c2009e881e6c395ab470c1717735389ed9dc64aacb8fe57311cb1f459947d",
   "videos/view1.mp4": "1415b316a3cdd62bfef246227ae529242e60adf53d1c2dcd97fbe1d8adad84ce",
diff --git a/Simul-Robot/c001_episode_002900/source/cameras.json b/Simul-Robot/c001_episode_002900/source/cameras.json
index 529b41bc2c36e1ba8f72adc86b05eac15a254c05..a365d84493a6db38abfc827208e92ccc537abb6d 100644
--- a/Simul-Robot/c001_episode_002900/source/cameras.json
+++ b/Simul-Robot/c001_episode_002900/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002900/source/meta.json b/Simul-Robot/c001_episode_002900/source/meta.json
index f0ad782b7297215cc5d9d4da8bb0de4f95fe09db..96af96306d0a6c817f4c72e170b33a60ccb92366 100644
--- a/Simul-Robot/c001_episode_002900/source/meta.json
+++ b/Simul-Robot/c001_episode_002900/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002900/source/visibility.json b/Simul-Robot/c001_episode_002900/source/visibility.json
index 4649237fc7c5ca57a4f87d727b177e1811ae2694..09c1dd806a9eb6c4f6f2e5191a40daf41ec779ad 100644
--- a/Simul-Robot/c001_episode_002900/source/visibility.json
+++ b/Simul-Robot/c001_episode_002900/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "tomato_soup_can": [
@@ -5260,9 +5260,9 @@
       "target": "tomato_soup_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -8750,5 +8750,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_002970/sample.json b/Simul-Robot/c001_episode_002970/sample.json
index 2cca8cea8d4331b9518b611ea9d4594b6d3df0e2..32e204114babb3de78b0e60ee5b7ca40e5bb0b53 100644
--- a/Simul-Robot/c001_episode_002970/sample.json
+++ b/Simul-Robot/c001_episode_002970/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "6abdcf1c2ada69bcb392d6fe9d5dde766d471c25600ccbc262f8739f5bf380da",
   "labels/objects.json": "443f0fa7f0488e2d74637d9c98e21f3d6a20cf72d44bede08c1dbc8299197a6f",
   "labels/visibility_states.json": "650669c9f29ba1bfcc31f3c48a7392f33be4ba0d6f8a7b4a5cca2eec4493f178",
-  "source/cameras.json": "0fb498fe5dcea7bcfdda0452c0d97e1be6be77094571539dc88c184ecf37b77e",
+  "source/cameras.json": "35d8688fed245134c8b22a36f631915c1e44cf5b63d29c7137fa824bf8b429fd",
   "source/caption.txt": "486f0a0a950bde33ca5a48eda25cae0085870d490eddb61fd9da8d0fc219bebb",
-  "source/meta.json": "72ca433c2abdd212253d227592707272dcff17c8bcb8fc8fb0480e6c461743b7",
-  "source/points_target.npz": "e8972b716498712a9addf2297eb88c4a01d4a5e7aae379b06503f3227e5379cc",
-  "source/visibility.json": "c7134597bd912d7b3fbc4a586768aa202ad6f070d8aa0eb17a418f00d2fc4f63",
+  "source/meta.json": "7856717430e43597f4e78de17628cac7e55cdf60b0653013153b6d66a0724903",
+  "source/points_target.npz": "1084faccc94c5630fe96b66773c913a3ccbb2fbc10e60a6777a103976fe90b30",
+  "source/visibility.json": "37361864f3b4e18dfbc1ee75dd72ef717ff7694ef71708c00100715fc77c8ace",
   "tracks3d.npz": "3772442a49c08e3e45a61b9acb7d2265a97406986ae51a394c3666bd9c7b6b6a",
   "videos/view0.mp4": "fc42c4875b1b45abcd667fe365eb8f092e294e4fae695f174597c04cf008e02b",
   "videos/view1.mp4": "5df3f1aea603d1706374aa3500cc0f533fa0a564ef026fe3c8f747dc55aca927",
diff --git a/Simul-Robot/c001_episode_002970/source/cameras.json b/Simul-Robot/c001_episode_002970/source/cameras.json
index 5edd929822149b07dab2fd8bb396c6bdde21fb33..30eff708ca2ed7766d2a859928f381a30a06bd4d 100644
--- a/Simul-Robot/c001_episode_002970/source/cameras.json
+++ b/Simul-Robot/c001_episode_002970/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_002970/source/meta.json b/Simul-Robot/c001_episode_002970/source/meta.json
index 21385a40eef92150d8375bbe1fae2a6eb8704a7f..3619b70fa5d6dcaad6a3943fc0a8b55e8afbad82 100644
--- a/Simul-Robot/c001_episode_002970/source/meta.json
+++ b/Simul-Robot/c001_episode_002970/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002970/source/visibility.json b/Simul-Robot/c001_episode_002970/source/visibility.json
index 61fd638ed23789ef6eb07b1b46ff704013879dac..e30cbaa6b4ad2f4cb38023d5e41e2e928e5eaca1 100644
--- a/Simul-Robot/c001_episode_002970/source/visibility.json
+++ b/Simul-Robot/c001_episode_002970/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "green_tuna_can": [
@@ -5980,9 +5980,9 @@
       "target": "green_tuna_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9950,5 +9950,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_003010/sample.json b/Simul-Robot/c001_episode_003010/sample.json
index 18ea68587c8354b5ed466c1f04a7e4702b412e12..bdeb0ef2f5261405dac1cd59740354b3dd3833c5 100644
--- a/Simul-Robot/c001_episode_003010/sample.json
+++ b/Simul-Robot/c001_episode_003010/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "e424dd26a5841f5f5d134d6170f7a7ab250786b40b2330b12a80e4ce8dece31c",
   "labels/objects.json": "befebeb55f569f9d50bb4d93c33bfc1dae7730d0688af7e498e463f337feea59",
   "labels/visibility_states.json": "cdf309991d3a41db3398a05f7e9ddff328922b79379f3033a24ce1122fde2a56",
-  "source/cameras.json": "a1153cb00ca79539f9819ee2e22bced5d93a7abf2d34fe1e27b4d21b42006802",
+  "source/cameras.json": "de01983f7abaf305687c3ec301ae98094f0e202877638c8b3cdb83231be195bf",
   "source/caption.txt": "68ce39236926a34e21b06d97cc2eddf63971892f1132e9d85c4faf2a69bf1656",
-  "source/meta.json": "6340c9f370d8a9a29a2468ce5b5f248331d73b4b89f2c843efe28e847141f1c0",
-  "source/points_target.npz": "f74a056a9b119768ce676f3ad911e6fc956252c8e5912bfb6937f09904b87d52",
-  "source/visibility.json": "8cabd1d33dbbe4f611a328f184785ac91c1a82ad40d81f465a69e2f73a9f30ea",
+  "source/meta.json": "2af779735f56a822aa5c7b17060a419e3441a225c83526bb1323f72acf7bb826",
+  "source/points_target.npz": "4268f12b85824d4c748d41c144ce63c09deff9aa66aae5aa90abc1d1781f6d1d",
+  "source/visibility.json": "88b917be6f880cf3f01d808958a2d15e749ff7d7cb5416fa28f072826eb75036",
   "tracks3d.npz": "a8c13273da5b814342b4b898a9706ff98b6842c52227cd6c6b8cda90220a7524",
   "videos/view0.mp4": "35ea57e1236bf50462003d0b78d4182b710657983d17756aced30c3335986dc0",
   "videos/view1.mp4": "0ba8c5ad16de652e240ec7ff06d44d5c489bd15367d28462f367801a446d3fe7",
diff --git a/Simul-Robot/c001_episode_003010/source/cameras.json b/Simul-Robot/c001_episode_003010/source/cameras.json
index 20de49adbaa8d02e72fa3c2719046574ffb88aa7..efeca44f4283ab8405c6f4b764870c496326c02c 100644
--- a/Simul-Robot/c001_episode_003010/source/cameras.json
+++ b/Simul-Robot/c001_episode_003010/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_003010/source/meta.json b/Simul-Robot/c001_episode_003010/source/meta.json
index 4d820aa3078f481dfef112c6d5a63c47758e866b..e36ea1aba705a0a64f7dde881d5fd660af0e73bb 100644
--- a/Simul-Robot/c001_episode_003010/source/meta.json
+++ b/Simul-Robot/c001_episode_003010/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_003010/source/visibility.json b/Simul-Robot/c001_episode_003010/source/visibility.json
index ab102d6cab23293b04ea9297e0be81457c47a81f..832202215086ca4a57d4ac394276ef457c681939 100644
--- a/Simul-Robot/c001_episode_003010/source/visibility.json
+++ b/Simul-Robot/c001_episode_003010/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_rimmed_plate": [
@@ -5890,9 +5890,9 @@
       "target": "tomato_ketchup_bottle"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9800,5 +9800,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_003140/sample.json b/Simul-Robot/c001_episode_003140/sample.json
index 964eef24f9e4ca7f7051eefe7d1cf5392c9317bc..763cf33efd775423fb44cb609bdaad9a8127a5e1 100644
--- a/Simul-Robot/c001_episode_003140/sample.json
+++ b/Simul-Robot/c001_episode_003140/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "b1a5cf593af06d76dad6ef6d3313b080dba736eb9d2be81ce3477de680e9e6c3",
   "labels/objects.json": "3e6484471018f3a91cca0de741607e0da1a9ec2504851275256ac152c48195eb",
   "labels/visibility_states.json": "91c44d895cf9ef794bc449955ed6f550ad3fb0697e733b03e62b1f9df3da8cff",
-  "source/cameras.json": "704223d421c35a750d9b3e4367382cbc67375e995dfa795465e710d569506c8d",
+  "source/cameras.json": "f197e5590d47200142815427ac69a226d2e7a96355d17b65289c155de130bc21",
   "source/caption.txt": "bf734b345fbef87912073b24deb46085aa2ee40ae3c22e92da194741931e0901",
-  "source/meta.json": "6186707250ef491ae3b611f8ff0f8f6042cccfab8cac724437f36ac0e69bcf22",
-  "source/points_target.npz": "f2e79b2ecd1acc52e6c54370b69aee22e551c527b2dfabde6b5e517e7e84de34",
-  "source/visibility.json": "da1c610eecff28e94971d97e349ee1ba9e73d94ffc690b31863bbb3a0b2b5041",
+  "source/meta.json": "4066b300b833b531eb7b63f8492f0ef34b37443e9ca67398291ea41c231c3e62",
+  "source/points_target.npz": "c342ad11b98bf59c9430c674bfe0bfa39631bd1d04971363b9531203b333954d",
+  "source/visibility.json": "0bc0fefc18422bdad1b2a7d28d7b0a6b5c3552ddd1181656fda15701ab4c97aa",
   "tracks3d.npz": "4330429876e286b4893533337cf0c0c5cacde4729e187150c34fcfcb8d08983c",
   "videos/view0.mp4": "053f2f479f76c4b4a7d99577eac656614e8e818691c95fb1b2c6bda8ac3d6a88",
   "videos/view1.mp4": "1289518219ca42fc65fea617b12aa08bad1c709c20b1e70f3b3f3a372fc0f695",
diff --git a/Simul-Robot/c001_episode_003140/source/cameras.json b/Simul-Robot/c001_episode_003140/source/cameras.json
index 5eeefc1e7043c16d5e99b6203f74943b0e9d6b82..5a548f58dda568ff2496b272dab2e12d5a2a1b51 100644
--- a/Simul-Robot/c001_episode_003140/source/cameras.json
+++ b/Simul-Robot/c001_episode_003140/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_003140/source/meta.json b/Simul-Robot/c001_episode_003140/source/meta.json
index 6ec77cb6f7a4fb649e9c38a0abb33d91e3cf59a2..b0c9cc2e7b29e2716c73defe7f10e7f7a7fe81c9 100644
--- a/Simul-Robot/c001_episode_003140/source/meta.json
+++ b/Simul-Robot/c001_episode_003140/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_003140/source/visibility.json b/Simul-Robot/c001_episode_003140/source/visibility.json
index 7f0de621287f725051a226a830e82092d946fcc0..8ec5f6633a6ecf132250a6637ff03c143ad65b23 100644
--- a/Simul-Robot/c001_episode_003140/source/visibility.json
+++ b/Simul-Robot/c001_episode_003140/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "bbq_sauce_bottle": [
@@ -6790,9 +6790,9 @@
       "target": "bbq_sauce_bottle"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -11300,5 +11300,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_003180/sample.json b/Simul-Robot/c001_episode_003180/sample.json
index 77e98b7f4e18ff56035cbffa6b170dba38e535fc..7ba15e2845629dc9eb7dbbc1192c258e65b96fe6 100644
--- a/Simul-Robot/c001_episode_003180/sample.json
+++ b/Simul-Robot/c001_episode_003180/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "ac3aa6104417ba530d97b79da12696f5eb8c20444d6776599adde95b6e31c191",
   "labels/objects.json": "2c981dc03627fc4bcfcf557f92136416fad9083de38b02dc0b8ae0d35f6fe2da",
   "labels/visibility_states.json": "cf75530140b907037daa6afd8a76e33d404279e3a02bd92a2b06acbb460f3bf1",
-  "source/cameras.json": "c08a72ac98a6ead98cb95cf10c647e4bb231e6867098225272d33b43d2fd54f4",
+  "source/cameras.json": "5cade01bb68dee6df2268219773f0b02c7fe793987d42e52df0ac7ee3b751fcc",
   "source/caption.txt": "1e332e455ae1cfebf9afb3e9bf51417a19615ad3f3d428c0a981df0096a8b5f1",
-  "source/meta.json": "d2b171eef0f06c2a9177a347f3d3c36a6919d1fe8d06d92fe534f376537f2823",
-  "source/points_target.npz": "3f0ddeb7c3fd7bd95fe386a18bdffee2b332922ca92bd5733bec2c6fbfb4f8a2",
-  "source/visibility.json": "c40eed12bda7b3fe627012dfc03113e9218dc5bfe55e19eb18add0cea4777bc3",
+  "source/meta.json": "0e24a1f050964e90ce301d41081ad4e06fe4ee208439c81c066ad293588d6224",
+  "source/points_target.npz": "e6640b9efce3b16575c7f92d8f4f0be9795b911eb19ca918c860054113e68f4c",
+  "source/visibility.json": "9135be0512236f5422c471203c950eea24af7b23a525ff2a81fba86fbf1fdfc3",
   "tracks3d.npz": "9ee8d2fbde745288c61dd0ecf03e5344f6ac22953b1096f136cccb5d4723932c",
   "videos/view0.mp4": "2feed2c51cccb635e3fb3a3e2460d5348071f782163364e8f6372c23c19fc840",
   "videos/view1.mp4": "754652efeb5a3f401d295daf135e68469fc71fba5a71641ee7a4755bfe39a69e",
diff --git a/Simul-Robot/c001_episode_003180/source/cameras.json b/Simul-Robot/c001_episode_003180/source/cameras.json
index b364bdf410ece546e03797c47275b60163c56c4e..b17b5873ddec5c26e2e4f899c24a8e8322bd41c3 100644
--- a/Simul-Robot/c001_episode_003180/source/cameras.json
+++ b/Simul-Robot/c001_episode_003180/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_003180/source/meta.json b/Simul-Robot/c001_episode_003180/source/meta.json
index b41e943e1d93246c4dd27a6286b6b81d5655579f..544102d62568185e5ddc16f59d0f4941dc889a5c 100644
--- a/Simul-Robot/c001_episode_003180/source/meta.json
+++ b/Simul-Robot/c001_episode_003180/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_003180/source/visibility.json b/Simul-Robot/c001_episode_003180/source/visibility.json
index 151045e5715e1215e6899c3be0d81ba578aa3257..fd63568f27a7667ba49c1cbbddf8d1f80745c978 100644
--- a/Simul-Robot/c001_episode_003180/source/visibility.json
+++ b/Simul-Robot/c001_episode_003180/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "sweet_corn_can": [
@@ -5200,9 +5200,9 @@
       "target": "sweet_corn_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -8650,5 +8650,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_003590/sample.json b/Simul-Robot/c001_episode_003590/sample.json
index 8ac1c68545eaabaa76b6d2bf18f9c85f911907c3..ddc65a8adfb96f6ba3b8016c31e2e97fe37f44eb 100644
--- a/Simul-Robot/c001_episode_003590/sample.json
+++ b/Simul-Robot/c001_episode_003590/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "a033a2cc05ecc5adbbce29859c94d900614394f76726971a98fcde498946e5ad",
   "labels/objects.json": "ce2d0345764be085e19d53306bf04debe0553133c363211f52c8408aa24ec17c",
   "labels/visibility_states.json": "39d46699de57995f841f41fbbece2392381a6cd7e95c65f0f5c41c232acb5fdd",
-  "source/cameras.json": "2add3558a27f81360b2ea119984752de8e1a43f35a6205fdc4348597100ace1e",
+  "source/cameras.json": "e5c528c573725e9692e682ec3dd0b3f7eec7ad7a8567a4258032a084ded6a0e2",
   "source/caption.txt": "f3a37c83e344b257906a94750a056f556b67ccdfb6265afedd81d24a26cb9f44",
-  "source/meta.json": "aa7503ae8f18d7d694dd9b75784b3cfc6b2b464693e6771550d7c8789f782000",
-  "source/points_target.npz": "25601f04161d7a0d1d4c23609382212b2246dbd52848bf6ac1b316b808a8f7f2",
-  "source/visibility.json": "6d8757cf384f5adbe5ab04743655ed4a8c89812d1ab05aa3c0db5776368b313f",
+  "source/meta.json": "83fd10003c227d513f47f19deea6d675891c89a2f894a3e1347698b9941c6423",
+  "source/points_target.npz": "34b9608ead9ddbc0e78d2acabd1ac6d2247bc6bd02bff6907efe182f89d07bfb",
+  "source/visibility.json": "d0a496b9079eac1f17ad275b63fec2c3c5a5b6c73abb137cd01ad24948ec3e24",
   "tracks3d.npz": "8f7af4659a04abc019cfea8b947593678baf2424a058a52783b8b586e4b62d31",
   "videos/view0.mp4": "0036cfeea9f7e92423f695eecd6f46b34f9ecf6354a9922c87f0619b1dc3533c",
   "videos/view1.mp4": "0df78472684b0bc468150c26b7c829d736752bf794ca8b1a1321efb346449fcd",
diff --git a/Simul-Robot/c001_episode_003590/source/cameras.json b/Simul-Robot/c001_episode_003590/source/cameras.json
index 2e4b99fc78088294eff8c0beaa565692b7f20d91..dce9075dbfe3dfd9cf0e01b6e51a47495aa92434 100644
--- a/Simul-Robot/c001_episode_003590/source/cameras.json
+++ b/Simul-Robot/c001_episode_003590/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_003590/source/meta.json b/Simul-Robot/c001_episode_003590/source/meta.json
index 4950520e5358df0f802a1ac53b19b307b5535985..24e4b9e3ddd1f9d230832ab7666ee59231983b7a 100644
--- a/Simul-Robot/c001_episode_003590/source/meta.json
+++ b/Simul-Robot/c001_episode_003590/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_003590/source/visibility.json b/Simul-Robot/c001_episode_003590/source/visibility.json
index 54d5b3006777c211a892f5a90dc309cd04a9d582..20a525ab608097be53a5d360b021702e8dc914c0 100644
--- a/Simul-Robot/c001_episode_003590/source/visibility.json
+++ b/Simul-Robot/c001_episode_003590/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "floral_cat_figurine": [
@@ -6340,9 +6340,9 @@
       "target": "floral_cat_figurine"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -10550,5 +10550,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_003650/sample.json b/Simul-Robot/c001_episode_003650/sample.json
index f94a3de9463e3420b5039463a2ad4ec028baa15c..782ba9dd302a57452eeacb5a4250898c5013e4b5 100644
--- a/Simul-Robot/c001_episode_003650/sample.json
+++ b/Simul-Robot/c001_episode_003650/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "0b33891602ec18b27d67c217d5c50f54b9d9ce6193157f063ed36d02d52abe35",
   "labels/objects.json": "121b27f5e0608c951d3c1fd1c9a72896c6568d90ccbec4f1a24f9d0e66404ca6",
   "labels/visibility_states.json": "86f712f2e9986aa7c233376c52657b41c45706019de577da988bb7bdcf9aac42",
-  "source/cameras.json": "bc1c1a1821a9f8ecd4b93038a06ead262f2907dd1f3fe758dc4d2fc3c8114b60",
+  "source/cameras.json": "0dfd6e39b333a60c187b178ce3cf19fdd60fa8c469c7aab7853b81da66c60e00",
   "source/caption.txt": "5f2e984f326bf528778438e31d046bc269fb6074b6e1d8a288f63a5933154083",
-  "source/meta.json": "2512619d81824a90997c183cc01b08410a5abf2516fb3db8eabfe8a089b64227",
-  "source/points_target.npz": "2069102e20f25d80c8e0ae99b2fdc00b8e760d61faf67502924c3aa55b3199eb",
-  "source/visibility.json": "237cc7ac5a15e1800e9e0a48897d9138a809a7a88dfc7e3d4d3c03bb027dc089",
+  "source/meta.json": "5120c3b83b9cb29307c70054ad8d53c14be3cb6adc26e703425b6e3d4e7fbd75",
+  "source/points_target.npz": "0d78496b92f41cbf798ae98da9d2a334c658cfb55d305208f4ebeffd4f256ad1",
+  "source/visibility.json": "a0d2d765fba201888cdcfc66a902a96b2221c972e4d34efdb912f36ea173d08d",
   "tracks3d.npz": "5fa0b8a527ac11f35cac000ad22e531ecca0be2792df2d0172fda23dc6d95a26",
   "videos/view0.mp4": "5e472e415641fb6abbeba39ca89b47d1537b526d6455fd62f2a9f256f0c6a1e5",
   "videos/view1.mp4": "409886fb9de8b6e372d379488e3db9fd4a48037cb7452187fa397e8445294c8f",
diff --git a/Simul-Robot/c001_episode_003650/source/cameras.json b/Simul-Robot/c001_episode_003650/source/cameras.json
index 1159504bc7bf8f5f3fe15a55432ecdcb32c325b1..34a8653fd6336d230cd163c5f39ae68fbe0af881 100644
--- a/Simul-Robot/c001_episode_003650/source/cameras.json
+++ b/Simul-Robot/c001_episode_003650/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_003650/source/meta.json b/Simul-Robot/c001_episode_003650/source/meta.json
index 8c1941439961c4a07136e683e2fcfcaac49f21c7..d809c57e073a0a2793070430c13d51c6814dc509 100644
--- a/Simul-Robot/c001_episode_003650/source/meta.json
+++ b/Simul-Robot/c001_episode_003650/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [2, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 2f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [2, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 2 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_003650/source/visibility.json b/Simul-Robot/c001_episode_003650/source/visibility.json
index b7d0d4e8d753eb20704b1a475d07ab628ab915d6..d8333fd81a353aed05d788f8b13673043d6c3904 100644
--- a/Simul-Robot/c001_episode_003650/source/visibility.json
+++ b/Simul-Robot/c001_episode_003650/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "art_deco_bust": [
@@ -5233,9 +5233,9 @@
       "target": "art_deco_bust"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [2, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 2f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [2, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 2 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -8703,5 +8703,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_004120/sample.json b/Simul-Robot/c001_episode_004120/sample.json
index 9993f7497ed4b714ed8de81c481473d7d4de7f2f..3d561334373302a82e8629afd89a242a4de87886 100644
--- a/Simul-Robot/c001_episode_004120/sample.json
+++ b/Simul-Robot/c001_episode_004120/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "8740b61cada83ddb9922064ed25e76f0067a2fd4d613e3b7b679d77b90c6cd47",
   "labels/objects.json": "434ac59e8614fbe0d3643c94de4e3aff040a4b60bca5aafe524cd71827815222",
   "labels/visibility_states.json": "e822bce6adb5926b9c3c989f2ef2ff5e5f45c461d733c98f0d2a34496821d0b1",
-  "source/cameras.json": "d942260dfdf2d2ffd4cc36d56a0dc2281e13e459006ff577016d6b8d5283b697",
+  "source/cameras.json": "071afd18a275866462053db9d9577aebcebe02433a918c21f90f82e8187a459b",
   "source/caption.txt": "c4fc40b7e2d368abf6a86896a87463dd839da240c32c15fd1ac9ed2ce396041f",
-  "source/meta.json": "b790c407c76a498344b46a2280daef9a3d39c5b934335b6150ea5f394e31c628",
-  "source/points_target.npz": "3aac3a7ae495b4a61b68fb6d7105e4a3b4e78b5181c777661a11075ac802c4e9",
-  "source/visibility.json": "777c5c1de3b842ea221f41bb1776a4cd05c40bd6ca7597ca6cbd3a46fec201ed",
+  "source/meta.json": "c300e9ff583e809ee730b8594762d3e3bbfc37df94564bcad06bd72ebf9da40c",
+  "source/points_target.npz": "fb66adf698a936e5bda7a8891d1bfa1bbc62344d53cc3b40145da25664d20337",
+  "source/visibility.json": "150a95bd2e6d5cc0d8e2358dd4d1e76f5d4eb9e501a8595652cbdbd44be15709",
   "tracks3d.npz": "02ffe28a12c86b0614370a4087adc3b91c04ebcd9c7b3581447b94f1ebff797e",
   "videos/view0.mp4": "f9ec516b7cca77eb9a11a3afc6501c1fa887e4e8fcd7acbdc671acddceec1bb2",
   "videos/view1.mp4": "5a7c4071788a75b2d763da2908d4099cdd0b12bbf1864c120b95ee96691387e9",
diff --git a/Simul-Robot/c001_episode_004120/source/cameras.json b/Simul-Robot/c001_episode_004120/source/cameras.json
index 55361cb77997685b942c289bb56106751b9056c7..0996195061691bd4f6a3120b0c73f3435c08e1e4 100644
--- a/Simul-Robot/c001_episode_004120/source/cameras.json
+++ b/Simul-Robot/c001_episode_004120/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_004120/source/meta.json b/Simul-Robot/c001_episode_004120/source/meta.json
index 14e5da350ba0f9663df736fbad57fadf26b44fe0..6b69ceb8246c8f669dc1ea963aa4efcc36da42bb 100644
--- a/Simul-Robot/c001_episode_004120/source/meta.json
+++ b/Simul-Robot/c001_episode_004120/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_004120/source/visibility.json b/Simul-Robot/c001_episode_004120/source/visibility.json
index 3e590b73d8381378fdb2682b59cbd3f2a6c71e89..2afc743c925e10397433d8d1bcbc35d5f38dc822 100644
--- a/Simul-Robot/c001_episode_004120/source/visibility.json
+++ b/Simul-Robot/c001_episode_004120/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_frog_can": [
@@ -5980,9 +5980,9 @@
       "target": "blue_frog_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9950,5 +9950,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_004510/sample.json b/Simul-Robot/c001_episode_004510/sample.json
index 45cdb3e123583177b44e3075c7e36ee0acc98d4c..e05c4f63f928884b319d771f89286f6fc11605d5 100644
--- a/Simul-Robot/c001_episode_004510/sample.json
+++ b/Simul-Robot/c001_episode_004510/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "4465ef8f2d5a2b888275418ffe92c379d978b5f7ed69f705bb1afd1f2203cfb9",
   "labels/objects.json": "98c1025a4c3e12f75b9ef7a0814e04df55f6d3cac33a4950b5595c205d1a43ff",
   "labels/visibility_states.json": "73b8aaf5e8384018e7b7259e7684adfae27ad362d74aef306a91f0318f228ecd",
-  "source/cameras.json": "593131bda2b1cc3b451288e983224b48478eb54ad3a54a3256b133c48a3c721a",
+  "source/cameras.json": "a3aebb3a3d88c2f83a7b8da6614ab595ab3f5dc0703aa20ba3012e0f92f7ad7c",
   "source/caption.txt": "ff32a8004840e2ac50a4ebc00d4387e28c34b514d061dc33ecb7d40ac04c536b",
-  "source/meta.json": "be652f480724ee739c65e37f6341b1e360a2fe947489765b5e123b086ab0f126",
-  "source/points_target.npz": "63865241849a5562054694f0764ac1739a5d9ec8649f8b6139f06d78076d49c9",
-  "source/visibility.json": "794502e966f98f648a0d50b651168fd22aca533c0a058ca0728d4456efa47b65",
+  "source/meta.json": "b201f5bf75213d83cfbfc25fbcecdd1f7e8fe5c19604983e88ca57407c46edaf",
+  "source/points_target.npz": "02917ffc3ef5fa34d5491cfa7f299b8649da4ef5d37a9bf3c7f1a8b5f7c08bf7",
+  "source/visibility.json": "4c3dce62ea128fc2f1d4d5098db5e6a4d4ff5ea214e877d7db65b0e10301bf72",
   "tracks3d.npz": "80b955e077d10a3be5445855f471853217158d0d631a1e5a5e49b3656973c84e",
   "videos/view0.mp4": "b70a27ddd3b702251f6b8fd5e5192a10d5d77eed97c848dcff2138e445b9368a",
   "videos/view1.mp4": "364af2d4f36879fa532d8276fc23998a27147bc5525bcbe6226f9e71c36224ae",
diff --git a/Simul-Robot/c001_episode_004510/source/cameras.json b/Simul-Robot/c001_episode_004510/source/cameras.json
index ca27d7a7ac0aef168a431da86ae9f0cef7a090a0..2bcd671eeb6430f450aaa3624e22d99022def44f 100644
--- a/Simul-Robot/c001_episode_004510/source/cameras.json
+++ b/Simul-Robot/c001_episode_004510/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_004510/source/meta.json b/Simul-Robot/c001_episode_004510/source/meta.json
index 2c7ba278b30dbe94a89e0dbe5ef7a9d56d6e5537..4fc75898d73f09cfd52408f9034cf050226a6f3b 100644
--- a/Simul-Robot/c001_episode_004510/source/meta.json
+++ b/Simul-Robot/c001_episode_004510/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [4, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 5f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [4, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 5 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_004510/source/visibility.json b/Simul-Robot/c001_episode_004510/source/visibility.json
index fb041a6a5a2c5014f299764e372285540cb35693..42e43f123568eb54d05409d11db11bb0c6deebea 100644
--- a/Simul-Robot/c001_episode_004510/source/visibility.json
+++ b/Simul-Robot/c001_episode_004510/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_milk_carton": [
@@ -3733,9 +3733,9 @@
       "target": "red_milk_carton"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [4, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 5f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [4, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 5 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -6203,5 +6203,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_004520/sample.json b/Simul-Robot/c001_episode_004520/sample.json
index f4002d652bfb49a57e4cc87160d4b18d8dfa2c9c..86358d22aa4f148ea8b6307e8aab8f6be79350a0 100644
--- a/Simul-Robot/c001_episode_004520/sample.json
+++ b/Simul-Robot/c001_episode_004520/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "e09c365023f6eedddd75ac8722b69543bb3f906ea26b7a7c241ff7fe197f43b6",
   "labels/objects.json": "03a6c125fd4607fddc5750625cff864e2507b9d3099799d39f0499c578884b82",
   "labels/visibility_states.json": "ff6826715d6fe4194c02156b90062a27b82ff9bbde2c3be286f19aca7b904539",
-  "source/cameras.json": "cb14f198e232cb43f569f7c2e01c11bc6bca1068fae1b2affc5036101e23cfb4",
+  "source/cameras.json": "11843374a6b2f062595c3d997748a15f36352b353750d2c1099643b4be6b191b",
   "source/caption.txt": "2261946efed76df5ba42ee61f7796b57da2ec85dbc90f77c0a9ee5d62aae3dc1",
-  "source/meta.json": "c1e7ff084d353e9550b6a59236d272b8689df3f49a48642d7ea9ae27f3195e5e",
-  "source/points_target.npz": "c84091ea1a2eadb33bb6d9dde5865779f06fb16f0d3389b89b9e94b179fc1a3a",
-  "source/visibility.json": "943af226eb6f392601034e4daf1ed29a29755a56c65b2a845e105dcfc88b4f50",
+  "source/meta.json": "a85314477484be1faac700e2fe8840ecd9b01744145bdaa5c96d844386d451c9",
+  "source/points_target.npz": "c210d97df7563243cd09025838caef6ec0008653b00b70b02a0e8d1b3e973fca",
+  "source/visibility.json": "bfbd12bc6955e394cd6d6d092a04d52a5248cd96683eccccc2ea8564b98a8bd2",
   "tracks3d.npz": "b71640a201e700b13d23f57c283aaa944e94a6c2e199e9a609bf05a9d7431153",
   "videos/view0.mp4": "18503dae7d716f6b4c78c7fd906d422364ad1022f5da8758d9476e7b0a1574f8",
   "videos/view1.mp4": "e7983be93fe3e728a6bb6b4991ec888961210a6b1948b3ce755ac5c65513f1f4",
diff --git a/Simul-Robot/c001_episode_004520/source/cameras.json b/Simul-Robot/c001_episode_004520/source/cameras.json
index 017aa27d99c79d87f8c9d4e1b3c38084763882ab..49e10f95a57e09f160abd35ff7d5f6ffb1a4c62f 100644
--- a/Simul-Robot/c001_episode_004520/source/cameras.json
+++ b/Simul-Robot/c001_episode_004520/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_004520/source/meta.json b/Simul-Robot/c001_episode_004520/source/meta.json
index f2dfe8c1f350ff47e2f7313414c46f7c79fe108a..a1c2857f54692f59ccf20bebe787f810bb4cba42 100644
--- a/Simul-Robot/c001_episode_004520/source/meta.json
+++ b/Simul-Robot/c001_episode_004520/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [7, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.0, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 7f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [7, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.0, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 7 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_004520/source/visibility.json b/Simul-Robot/c001_episode_004520/source/visibility.json
index 2a727af1a55be8d9b31b0ecc1ee5c0dcce41bd47..789b7521cc61f930397ae39fbbf3784a12bf9698 100644
--- a/Simul-Robot/c001_episode_004520/source/visibility.json
+++ b/Simul-Robot/c001_episode_004520/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_frog_can": [
@@ -5143,9 +5143,9 @@
       "target": "blue_frog_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [7, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.0, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 7f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [7, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.0, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 7 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -8553,5 +8553,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_004530/sample.json b/Simul-Robot/c001_episode_004530/sample.json
index d8b4a4f632fd27ac40964a84c661a808402e5216..575d362fe26bb6b74e4eeca9fb54955f50fdd5c5 100644
--- a/Simul-Robot/c001_episode_004530/sample.json
+++ b/Simul-Robot/c001_episode_004530/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "a7e14f706a655042bb832479d8b28d1921f935da67537f9ad82a947b5ed37dc9",
   "labels/objects.json": "20ef9ee24e79a397b696d2589de66826129bfc6afc4f9929ff1f01bde3f98e33",
   "labels/visibility_states.json": "cf9cc649254259bd0d95d04439a4f0762332d713964f2f9f6fd4d9388ebbd55c",
-  "source/cameras.json": "6b7a132525b7e12371c1bc8c6d541ec9e9d121baf18dc73afac484c6f2086573",
+  "source/cameras.json": "0e92e83d239debfcbe5637103761d3a38591c80326aaad786cfe75dfbb3db3f6",
   "source/caption.txt": "8151fd2480a837eaf4dc4987780cf0afde50274f23d0fdb810fd552dffe242fa",
-  "source/meta.json": "8f1fb46e496cfd3cc8ddaa58cafe3f940ade3a37e0159cbc3973d7112cf9e844",
-  "source/points_target.npz": "a605f05d894b0dd5a101943beab05ade104a2ba5b4dc4b1b53b316ef9663e404",
-  "source/visibility.json": "9dd7434d5182f98ac3016c628482754e378c4258270bf804795f0b6af059e09c",
+  "source/meta.json": "d7e5749a425b14fe12ce99b74b52ecfb223a0edc48e580f4150c35f342f8db9f",
+  "source/points_target.npz": "f7de8dc68ca0cf582ee0170fb9b7da8a36567d480a79349f51130ec39e25556b",
+  "source/visibility.json": "b1440c8205e6302a9e51874f4d2b284613434be6972931f86c13d9f8588ed42f",
   "tracks3d.npz": "d1d06e78ff057b50292f790db52246e7b115578f1195cd15e9b4144cca369664",
   "videos/view0.mp4": "5834d2b4f15118bf4ac89de212fa0758f2b87156705dca62dd5b78d3a41cd755",
   "videos/view1.mp4": "6020c39ac95469bc113312d22dd7027dbcc958ec6b861cffa77351a0a98a2017",
diff --git a/Simul-Robot/c001_episode_004530/source/cameras.json b/Simul-Robot/c001_episode_004530/source/cameras.json
index 9cfc2944be75a8c7a9c1b7b6bfc5153aa8cd219f..b28c02dff3ebdf1d8786f670f11a1f455142828a 100644
--- a/Simul-Robot/c001_episode_004530/source/cameras.json
+++ b/Simul-Robot/c001_episode_004530/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_004530/source/meta.json b/Simul-Robot/c001_episode_004530/source/meta.json
index f49aacde4bc5fd31010307182a3a6fb98c89c6b1..14f43f85bdc8de75ec3a03157d22a6bc68fd44bc 100644
--- a/Simul-Robot/c001_episode_004530/source/meta.json
+++ b/Simul-Robot/c001_episode_004530/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 2,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 1]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 4.16] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 1] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 4.16] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_004530/source/visibility.json b/Simul-Robot/c001_episode_004530/source/visibility.json
index 71b632f5a0b3206b8a3c310fdc1253f0a87eabce..114eed7d8f53b8c82801dc429e6ce89f66ac051a 100644
--- a/Simul-Robot/c001_episode_004530/source/visibility.json
+++ b/Simul-Robot/c001_episode_004530/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_rimmed_plate": [
@@ -5596,8 +5596,8 @@
       "target": "tomato_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 1]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 4.16] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 1] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 4.16] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9305,5 +9305,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_004540/sample.json b/Simul-Robot/c001_episode_004540/sample.json
index 955209ad257bcbfb175b7f0cc8713fe770290e7e..a5d2402598c3e0413abc3fb109edbf579dbb4b3a 100644
--- a/Simul-Robot/c001_episode_004540/sample.json
+++ b/Simul-Robot/c001_episode_004540/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "79e56572d98804e4579f28e5531f19c199de301f1723863c9ed87842a2a203e2",
   "labels/objects.json": "38ba0f6bd728be61a7c718a8e44150c96adac7ff0a53509ea12119479c1fd1dc",
   "labels/visibility_states.json": "149da193433cfdf3cdb267c684bb794a5df691d1f28779f27837bee9f1227c5e",
-  "source/cameras.json": "c81902dda5238c6d76d5f6cebc3545ce8148285f2f4c146ee12cbf10cac772c4",
+  "source/cameras.json": "d723cae176379cf5154ef40537dd9ad4e8aa0c335136a4e888166702101f35d4",
   "source/caption.txt": "28953ec5ca37579aa9f871ee7c1c637f0f9164053795913b7fab98d7475f5ec7",
-  "source/meta.json": "d101d76e8c303ceb618ac9cda89343902d2a0151ec31e27e0d827c4cec20a000",
-  "source/points_target.npz": "ff2062807ab7411e878150219115a18d30d51e3ace2a6311180a2de6a7097a3f",
-  "source/visibility.json": "e5e576d1220dd26faa1dec71954d8bd7fe2b5e511a8482226c6eb172bfbaf7d5",
+  "source/meta.json": "afbd1da37f23cf0122d1897768e4a6ff802dbaca321ee1bf55c8534da50a648c",
+  "source/points_target.npz": "eb5b02a15861a4a7cc2c158f0b0aa52234b3fd21ee099a2c3a583bda106b16b6",
+  "source/visibility.json": "f459c71321f30c022e4bf151b8658fc89037c69da2a514367acea850d57993e2",
   "tracks3d.npz": "28f2d3ace380ffd5ae0d9c737b8fdb57b0b07f03cf7d0d17731360152d0a9f66",
   "videos/view0.mp4": "aaa9b76eb7b476ec449130d2f935365dfd4302977604c0237804ef6b7338cc23",
   "videos/view1.mp4": "a98a74fc08b21a735c870b88229718e399862cba34450c613ff39d7c40eafe13",
diff --git a/Simul-Robot/c001_episode_004540/source/cameras.json b/Simul-Robot/c001_episode_004540/source/cameras.json
index 1ec3e553544d34273da4d77409b34860502abbc7..f47c5495acb7e5ccded602f9754fb25c515dbd77 100644
--- a/Simul-Robot/c001_episode_004540/source/cameras.json
+++ b/Simul-Robot/c001_episode_004540/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_004540/source/meta.json b/Simul-Robot/c001_episode_004540/source/meta.json
index d1376ced34487ad255b57a3e84ee114ef6652bd6..ca48855243e5fbb83f48946d59afb653565affc8 100644
--- a/Simul-Robot/c001_episode_004540/source/meta.json
+++ b/Simul-Robot/c001_episode_004540/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_004540/source/visibility.json b/Simul-Robot/c001_episode_004540/source/visibility.json
index bee095fd5cfb4cd37e1a5404edfeb29ed65811cc..71249db0b00a6ebf6c4df21ae58c4976c7588d14 100644
--- a/Simul-Robot/c001_episode_004540/source/visibility.json
+++ b/Simul-Robot/c001_episode_004540/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_milk_carton": [
@@ -6460,9 +6460,9 @@
       "target": "red_milk_carton"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -10750,5 +10750,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_004590/sample.json b/Simul-Robot/c001_episode_004590/sample.json
index c93817e37cdfc2807e4aeb5b4fea43ac608ddad1..a65255d688ecf59388dd41bfab2d8d1c087b00f3 100644
--- a/Simul-Robot/c001_episode_004590/sample.json
+++ b/Simul-Robot/c001_episode_004590/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "0aaf2ac67c360b4bf8c5b8497ff14af457371df5c5d10fef7e3c76459943c2fe",
   "labels/objects.json": "80b6a07fae231118c1451a846b66e16ccad2e82aa301beeb4f35260ec3d06ebe",
   "labels/visibility_states.json": "724a69f324f9096b46c427926d1b256401620ac5a0e4c105a4c0c5c99bb3228f",
-  "source/cameras.json": "c19cdbf797983b9b0326532dd01ec3b89cca2b6ff95685f0879bad02e6fd8c6d",
+  "source/cameras.json": "e6127187dd59fbfbe9b831ca76e5d62335d6dc24e5f2e1018860604e49b7a8f6",
   "source/caption.txt": "6098f1419124bc6026dd0c60a40d895041fc274fb1834aa4915ddf5d4d315f16",
-  "source/meta.json": "c54410615c110967b21c3628ad46974d28f0de6f38fe630055e34e5a9d21361e",
-  "source/points_target.npz": "b42fc0a1a361c1b12dd67af920b6f224e6f9cc71c1fafeef6af9c78d1e538358",
-  "source/visibility.json": "ffd996fcc4718719be59895e1d42bc7b56ac14e7af05e4cc34236d73b88aae03",
+  "source/meta.json": "a3aa270029e72ae56956c793bb701799fccfd3cc8194d1909d83e413e2847802",
+  "source/points_target.npz": "10a22d13788d241f910945cbeef12caa8030071d02c52b2b2d0e8b38cf2e0d89",
+  "source/visibility.json": "065f48411b8a0f2d3b525bf289c3d4ab80a9ee84319fad37237bc41bad93626e",
   "tracks3d.npz": "9472b245ebcb819c9a2dc341de7516d608caa681852050b1a3b34e70de1c3d34",
   "videos/view0.mp4": "93156eccc7ae332164a8b8284e35e00231b70cf6b9422ab033de46879e4db372",
   "videos/view1.mp4": "8b8725d4bf1f10464f40c86df1e3d2eb0b5c0c21988bc9e911a484d758b1eaf6",
diff --git a/Simul-Robot/c001_episode_004590/source/cameras.json b/Simul-Robot/c001_episode_004590/source/cameras.json
index 1bf188cf81c3362547d44992ed614b57f69035c6..947eb234d7b985ec03b4ffb82007e15aea4c4f77 100644
--- a/Simul-Robot/c001_episode_004590/source/cameras.json
+++ b/Simul-Robot/c001_episode_004590/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_004590/source/meta.json b/Simul-Robot/c001_episode_004590/source/meta.json
index 339f44a8f485304fc2ff09739f57dd4c07912de5..8119e9cec1505cb15d0c52f6b7f3bb19186179a0 100644
--- a/Simul-Robot/c001_episode_004590/source/meta.json
+++ b/Simul-Robot/c001_episode_004590/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_004590/source/visibility.json b/Simul-Robot/c001_episode_004590/source/visibility.json
index 9c8b9a67d375b4030c4c5edc6f41b6dfc725ff51..9153d590f1ddc2d5f211d91ed6ba4a15d5b6d80b 100644
--- a/Simul-Robot/c001_episode_004590/source/visibility.json
+++ b/Simul-Robot/c001_episode_004590/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "green_energy_drink": [
@@ -5890,9 +5890,9 @@
       "target": "green_energy_drink"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9800,5 +9800,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_004600/source/cameras.json b/Simul-Robot/c001_episode_004600/source/cameras.json
index fa05a006ce1e71d4ac9169100e65f10ceeba10d4..5b9845df2d3e1581639c5a0f472f84aecc572cd4 100644
--- a/Simul-Robot/c001_episode_004600/source/cameras.json
+++ b/Simul-Robot/c001_episode_004600/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_004600/source/meta.json b/Simul-Robot/c001_episode_004600/source/meta.json
index 3c998ec7a5d951a393071777b396a49170f6dd74..f4acc096bb5091c1cc3c4a3fd98de81d97403a58 100644
--- a/Simul-Robot/c001_episode_004600/source/meta.json
+++ b/Simul-Robot/c001_episode_004600/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [10, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.03, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [10, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.03, None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_004650/sample.json b/Simul-Robot/c001_episode_004650/sample.json
index e70826f7c81457a08a21b45c3bbc575154f43f84..c8e6aa2607d05a07d6aa0e7cee2ae10583e2db81 100644
--- a/Simul-Robot/c001_episode_004650/sample.json
+++ b/Simul-Robot/c001_episode_004650/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "6de8b5501f42276927aba37aa215b7b004f13e44185d15fd8d8df9dd260dcbcc",
   "labels/objects.json": "213c65cfa9ed25a6279fb6bb5e11525bdf0c375fd85951604d286f172524d7af",
   "labels/visibility_states.json": "ef83d32a7d1ed816578df90f004c2de51f21e4d7259d086069aecceaead3ad6a",
-  "source/cameras.json": "926df3d5b234e2360a6e155dc6be9401af0ea33f949acc54b8327bab14e48dd0",
+  "source/cameras.json": "649db9cd45bc8750c9e03e38432af0e35ec97be4f1d42f96b652b2261016ba64",
   "source/caption.txt": "cecc5a732c59a0df3836596baa770b66fc71f1657ff8536582529ef70aaf025b",
-  "source/meta.json": "92d269928cafc6f3d062fd0da588ccce7f6e370de5f25581760a50ea6cbde037",
-  "source/points_target.npz": "0bf007e118089557124ec84889dfaa25dea941547c9a1ece67c6b0469a9738fa",
-  "source/visibility.json": "311975b3659ea4e81080acaa2fe1bde609dd8f1600f56478bba49cb983491d2b",
+  "source/meta.json": "2850672e48ee604ac4c792b8caa5884d66ad41f4cc60d074b6173451e6d02c80",
+  "source/points_target.npz": "7318d4efdb63566e4606b116366d266494eadf9b0916685166afdf6deeb27aca",
+  "source/visibility.json": "47bd0287acf104fba437f219ad926f249b4b0fc7b16a0d25cee797a649a03dbb",
   "tracks3d.npz": "50b36e6eeb7c22cb6a3c0bd947c3aa8dd89a2dbd0d62f68f4709a46db9b97515",
   "videos/view0.mp4": "7dc778b4f4447c1c8bbb7e874a14d68ef00bda1456c69a6e022c8ae16ff314f7",
   "videos/view1.mp4": "ae9b071c05c50d4a97cec35edb13e6f84e5d556c335e6daa6fe0c51fc92321eb",
diff --git a/Simul-Robot/c001_episode_004650/source/cameras.json b/Simul-Robot/c001_episode_004650/source/cameras.json
index 881d709db09671dc39c6a6be27e9483f04a09c01..fbdb221f5499418ffe3c9ba244e41a411fb424c7 100644
--- a/Simul-Robot/c001_episode_004650/source/cameras.json
+++ b/Simul-Robot/c001_episode_004650/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_004650/source/meta.json b/Simul-Robot/c001_episode_004650/source/meta.json
index 3cfac92f1711901218ae0e1d375a389221a6cf54..42dd8cca3a21e048bc53d6c8a3763e638969ebdf 100644
--- a/Simul-Robot/c001_episode_004650/source/meta.json
+++ b/Simul-Robot/c001_episode_004650/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_004650/source/visibility.json b/Simul-Robot/c001_episode_004650/source/visibility.json
index 4a1dbba36712c7026f05f24959ec90303fe072c1..6e3c16e30ea06766c8fdacae35d12bfc1cb1def5 100644
--- a/Simul-Robot/c001_episode_004650/source/visibility.json
+++ b/Simul-Robot/c001_episode_004650/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "glazed_squirrel_figurine": [
@@ -5230,9 +5230,9 @@
       "target": "glazed_squirrel_figurine"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -8700,5 +8700,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_004690/sample.json b/Simul-Robot/c001_episode_004690/sample.json
index 3245f15c97c41db3682a702ac70c75ed9ce90017..05dfad0f833b5df56481af7f7de61bc276ad5568 100644
--- a/Simul-Robot/c001_episode_004690/sample.json
+++ b/Simul-Robot/c001_episode_004690/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "59341d46f9eb1d71e9edd23780dc2b59508192ed8bc7005c6759ed8825febe50",
   "labels/objects.json": "b70cd517038719354cb779058f1cf66774a4801d6c1003a42cae030a964fab27",
   "labels/visibility_states.json": "99d15fc4e86fbf1af02d786129de94c7e425497186d0d0374d6740bb4aef287c",
-  "source/cameras.json": "f33e13ea56be1efa2a559d2ac03dc3010ed1e229c28e313dafb5648eba6aecb6",
+  "source/cameras.json": "1f14a7434ac33d2bf186c877920c5f98c8df57a1aed386df261c3c0ef3e7e588",
   "source/caption.txt": "1a016c7a665ad5dbed8efb6f501047cf4cc72a5571724f5d0c7be8f24385f760",
-  "source/meta.json": "7638ed912d136ab148f5e025780d84dc32231fd3e0f15604da906063e5e95fa7",
-  "source/points_target.npz": "f6ab46cc2feb40b813eaf8ab71c1121eef3f6284ce867fae22ade579f701118d",
-  "source/visibility.json": "d11f7e130cb20f24d0636d600c70fd1468f210068a25e62b387e66e98563af3f",
+  "source/meta.json": "733c7dc552bed413137ccd6e168c06d436c1f9a71c8a05da6ed0d7072e3b62d0",
+  "source/points_target.npz": "829e7c638a9104233c1c9cd4aef0086f272cb1708ab146fac1e46158b091a5aa",
+  "source/visibility.json": "80cb6016ddde0b364762717efb03af4d71d1bd037dbadf2a2ad910b259fafc98",
   "tracks3d.npz": "693fefb60c44344342be9cf3ae32f684490b837a5684dcf4882b715291987c52",
   "videos/view0.mp4": "11534887fdb4695168deda04be490531c1b02763b3c096f82450f7138fc0485f",
   "videos/view1.mp4": "7fb7cec96ec9a0a728b7978dd998a2d12c909a6d201f3dcfc7959484084b3e6b",
diff --git a/Simul-Robot/c001_episode_004690/source/cameras.json b/Simul-Robot/c001_episode_004690/source/cameras.json
index 87dfd4d85d296c08d75a4e4057dd3a18565d265d..0f6d39f3df52ad25e2be28adb0ac87889fcbc6a7 100644
--- a/Simul-Robot/c001_episode_004690/source/cameras.json
+++ b/Simul-Robot/c001_episode_004690/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_004690/source/meta.json b/Simul-Robot/c001_episode_004690/source/meta.json
index cae6ba825f07d006b628b1f8c3ba0f232ca68f47..96560c07c86c3c050fbc22335cdc6b27d469977f 100644
--- a/Simul-Robot/c001_episode_004690/source/meta.json
+++ b/Simul-Robot/c001_episode_004690/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_004690/source/visibility.json b/Simul-Robot/c001_episode_004690/source/visibility.json
index 3d304f085933ab5435a137ad4ab52c734a1d6eb6..03697b554308eecda202b2a01eda9b21f878b0ca 100644
--- a/Simul-Robot/c001_episode_004690/source/visibility.json
+++ b/Simul-Robot/c001_episode_004690/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "floral_cat_figurine": [
@@ -5530,9 +5530,9 @@
       "target": "floral_cat_figurine"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9200,5 +9200,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_004730/sample.json b/Simul-Robot/c001_episode_004730/sample.json
index bbe78e77e451a549c47a92dea65d4ddea23e8d61..075173f2a914b8bd0b94b143aff9332d12a751d4 100644
--- a/Simul-Robot/c001_episode_004730/sample.json
+++ b/Simul-Robot/c001_episode_004730/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "d58534e41e0b42aca243b6c551345ffb4fc6c4735b11fb937c61c8b88f93b80c",
   "labels/objects.json": "e0e5e7ba59db3417cff0fb393f780eab96ed276a086735ef07f5892d3c6c5b5b",
   "labels/visibility_states.json": "5d619fa527d34047bb539e079fb877df053821c40c82e8be823906061d73b572",
-  "source/cameras.json": "a7a2ae346e0cb800683e4c2b708668623ccf859fc4aca293be8b891a20a1c77c",
+  "source/cameras.json": "33b923699a4c5dee6aa9854ea3123cc25e3f051956d7b5a1948d4f35d9675138",
   "source/caption.txt": "7a6a3936cfe65ba08afcb8090412d22bd9cb9fe35460f625a1600a87e2f623ed",
-  "source/meta.json": "1637b23bdfb78284ed5f0332ce1dee9218e883c20b3cf2788a52a050ff42e829",
-  "source/points_target.npz": "62d1f748bb9e7ed452c1bdc5aaededea78fb6c5bfb82dc85f8c31b882939c87b",
-  "source/visibility.json": "ba15fac3fd496aa97e82280063fabff5bc41d461a814f76cb2f297b948fa4ba4",
+  "source/meta.json": "32f4c45563d84ad43e60583dc001d3b21cbccdcb65b92a704ba7848c664d2628",
+  "source/points_target.npz": "1b61efe622519e65e4913578575dea51ee9b90b5f9d3d1b4c7868fc9a2c57e89",
+  "source/visibility.json": "78bd1bd8542567a20f32bdee740fcff1e38449415a6169406196b1aeb782b8ab",
   "tracks3d.npz": "e314757a317f805c35b5cb9a86fb90c5d052e76c01b358db7880dbfdd8eca126",
   "videos/view0.mp4": "746f943cd3a9604bb925c99fc363c969b7ebc950bfa72207c4013ea8b683ad35",
   "videos/view1.mp4": "c8860140a5b25944327df6412447224440d37c3e53308ac4cad486afc825d703",
diff --git a/Simul-Robot/c001_episode_004730/source/cameras.json b/Simul-Robot/c001_episode_004730/source/cameras.json
index dfcb8b009dbde55c208683e0503e581396dae757..f0fcf1b1db7690a056be3d3de6dcf4b521203e77 100644
--- a/Simul-Robot/c001_episode_004730/source/cameras.json
+++ b/Simul-Robot/c001_episode_004730/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_004730/source/meta.json b/Simul-Robot/c001_episode_004730/source/meta.json
index a0b29dc114a78726a1119b126aaec3365d1855a3..1b57d3bb0c8899129b53065220a65fbe6d75ff66 100644
--- a/Simul-Robot/c001_episode_004730/source/meta.json
+++ b/Simul-Robot/c001_episode_004730/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [2, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.0, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 2f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [2, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.0, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 2 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_004730/source/visibility.json b/Simul-Robot/c001_episode_004730/source/visibility.json
index 27d650cb4ecae78ef1128e12ad75ea2bb70fc7f5..69d160663808d75ae5173730f11a9ca1f449c6ba 100644
--- a/Simul-Robot/c001_episode_004730/source/visibility.json
+++ b/Simul-Robot/c001_episode_004730/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "green_energy_drink": [
@@ -5863,9 +5863,9 @@
       "target": "green_energy_drink"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [2, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.0, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 2f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [2, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.0, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 2 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9753,5 +9753,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_004770/sample.json b/Simul-Robot/c001_episode_004770/sample.json
index 35c21091d3d7bb0ae3356bbbcca3a256dccbaa07..c112d3313e2578e13cc9c371cabfc500f9edcc9a 100644
--- a/Simul-Robot/c001_episode_004770/sample.json
+++ b/Simul-Robot/c001_episode_004770/sample.json
@@ -222,11 +222,11 @@
   "depth/depth.zip": "53bfb97c8b641cb6895a76d1bd96caaeac81b18b23701b56eb76bf56b46c9bd2",
   "labels/objects.json": "caa10261ac16c1a33f67749bc5afe78d168bb514cafb4e60e00fe00e804f53e4",
   "labels/visibility_states.json": "9f337910ab8aa3fa461b5ed8d9732e29f8fa229e4b91bdc4cc15dd65d8aaefa5",
-  "source/cameras.json": "b9d239b9d94d3a400df39ea9324819e80f24ad67d4e139ca927af73c32fa9ba4",
+  "source/cameras.json": "fc00e4d869aded701d3b94f1e6e87974efc2ff3d84f41ab09bf8f575979c443a",
   "source/caption.txt": "bf5eba1b38493087ededcb323b1c1059e556edf5f80c6d512d175af736ed710b",
   "source/meta.json": "1630ae6ffb0d690ee39efb3b9d619336e9a289c10ced52570db8c2160627adaf",
-  "source/points_target.npz": "e82061d889ad25412e2ae0b3fb719b48e7864ce767feed433e10b343434fb400",
-  "source/visibility.json": "01f93544424966c58e81c2c05ee351db6742200cd66241bd5d84df459864bbee",
+  "source/points_target.npz": "27d5941d1646ffd086b3b3b06f430d7a06261857eea6dfbbfb08fa0ce1867ec9",
+  "source/visibility.json": "31a972d0ba37c504ca5831ad5f28127b741c37bb4f3f5bc848ebe57da6f10bbe",
   "tracks3d.npz": "d62be8d194057e42f1387779068032d4a9a5e36d5cfc91381be704f1dcad8126",
   "videos/view0.mp4": "9832ddcda146b9b1c694e6eec903815f02ffcb48d898b930021928e85f27d3fe",
   "videos/view1.mp4": "3bbccb173dfe71709fac41da7a8dad92443bd19f354de6a4e040d969e16c05a0",
diff --git a/Simul-Robot/c001_episode_004770/source/cameras.json b/Simul-Robot/c001_episode_004770/source/cameras.json
index 66827d804eb4d0c08ca196b2d5507af40b017af5..d4e0ead91483a7b4e89b2155381ef9e8a11b95b9 100644
--- a/Simul-Robot/c001_episode_004770/source/cameras.json
+++ b/Simul-Robot/c001_episode_004770/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_004770/source/visibility.json b/Simul-Robot/c001_episode_004770/source/visibility.json
index 9dd1708f19551605fa89b37117f64c63389f07c1..e861687031797eddf001fca3e07a380dc5c9786f 100644
--- a/Simul-Robot/c001_episode_004770/source/visibility.json
+++ b/Simul-Robot/c001_episode_004770/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_foam_brick": [
@@ -8945,5 +8945,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_004780/sample.json b/Simul-Robot/c001_episode_004780/sample.json
index e11438b795844e0dedd85ba04f322fe7c8cd1cd8..0f5ecafea85397abd095529993b8284094a59bd1 100644
--- a/Simul-Robot/c001_episode_004780/sample.json
+++ b/Simul-Robot/c001_episode_004780/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "d59461e16652d8cfd522ec2bd0438e0b81fd56d0e72ad4537e99f9f4b3b12181",
   "labels/objects.json": "c85138f34483bd76602de148e7868ee10aef439b2aafbc928c1a9f6c336a6d7d",
   "labels/visibility_states.json": "90686c5a05d39d0bf1ffb22393ce44d4ad8a4fcb8a006512f7d74740ef65f751",
-  "source/cameras.json": "0bc157a0552e766101e52234a08990c1586101203ed4579a508369927385bd92",
+  "source/cameras.json": "0a1322ca5e70459ce1dbee20467a170a1ab059b756b694b510ba9af23a13fb8f",
   "source/caption.txt": "af81eca9e8711fd081f28ee25856539d6d5d362769f0b8bdb236f9e79065b20d",
-  "source/meta.json": "d105743aee4ee882ccc6bc62c8d92d5dc6aa7ca53940fbe0a75bc73d3394ca40",
-  "source/points_target.npz": "4713da1064fed3b72494863bcae83bafb64daae8ed434a3f76673d2a33b6520c",
-  "source/visibility.json": "35b4704a7a6ce9c6e15977794243c5c539839b7e604f9db586aaccb0e543b10c",
+  "source/meta.json": "a6961c94b270e12537001ef5eaeff3cb274d6ca0385e5180675e10db59bf8f90",
+  "source/points_target.npz": "6f001f4480789244a2a8d8210e1a7deb4ea89589fa6e55c8d8a00d06c2672270",
+  "source/visibility.json": "8eaffe4f35387d082cfcd5d1377f94d13414e8c1ddc1303476f0aa69fa94891e",
   "tracks3d.npz": "51715cc13bcce7c2c5315a3982c60d4bf874d78dd6606345967333ff411c2e7d",
   "videos/view0.mp4": "f2ae110b446c66d38fc606c43672cdb0b9e35b30475180fbfcf71c931e2e9c7c",
   "videos/view1.mp4": "8ba924d3551c497f95fd577a689029a2272d00b2e9b5020577fb5e30ee5bd778",
diff --git a/Simul-Robot/c001_episode_004780/source/cameras.json b/Simul-Robot/c001_episode_004780/source/cameras.json
index ffa4a5aea26d3d28b47d49243e8f44f80c636d88..3367da26002e93ffc43779312329a697bed7df9a 100644
--- a/Simul-Robot/c001_episode_004780/source/cameras.json
+++ b/Simul-Robot/c001_episode_004780/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_004780/source/meta.json b/Simul-Robot/c001_episode_004780/source/meta.json
index 9dc7d057a9096d26b45508d9d705079a6ff6521d..4617f8495bce7a2e9922164146ae412aa2642eca 100644
--- a/Simul-Robot/c001_episode_004780/source/meta.json
+++ b/Simul-Robot/c001_episode_004780/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_004780/source/visibility.json b/Simul-Robot/c001_episode_004780/source/visibility.json
index f5938b01f8d8e170d8776e30d0dcbf4bfb2ac02e..178936c01047b83020cdc2a9848ae1c8edea77d3 100644
--- a/Simul-Robot/c001_episode_004780/source/visibility.json
+++ b/Simul-Robot/c001_episode_004780/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_seasoning_shaker": [
@@ -6460,9 +6460,9 @@
       "target": "blue_seasoning_shaker"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -10750,5 +10750,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_005100/sample.json b/Simul-Robot/c001_episode_005100/sample.json
index 2b16cff4e6f221f342da6e5803e870ae293e30d1..f03fe3ece457ff7700d3726296747827dc485936 100644
--- a/Simul-Robot/c001_episode_005100/sample.json
+++ b/Simul-Robot/c001_episode_005100/sample.json
@@ -218,11 +218,11 @@
   "depth/depth.zip": "fc43a3e5e8f2744ec8634c95cc01eca1c2931890d05a534d3924124835e274e0",
   "labels/objects.json": "1345e09d06925f4d1de95ac394655ff1b9dd4e8bc091d35cf03b754a097e7186",
   "labels/visibility_states.json": "44e818c2ce8a8a6d9980a881a41d79bee77d4018618220e0be70a96ac191bddd",
-  "source/cameras.json": "6a8545fd8400cb7deb1b5fbbc9f45336a900c6f597a3d679dff4ec50338515e7",
+  "source/cameras.json": "6b4d9d1e2a9937afcbce2244d1bbfbc0f0e35c315dd16d8c5eb485981489a565",
   "source/caption.txt": "a27f0b9e6f4b25dcc3aa63c7c89f20ef27b4def1a852fed48cc3760b377a582d",
-  "source/meta.json": "590de7e43c2ee99eb563311a0850432c20e03ef33c3e9007dff25e9702ea7316",
-  "source/points_target.npz": "1c9e5c6ee66ffb3cef46afb403ecc47c2d71f2e3b1bb7fd2c481aaa79b7588df",
-  "source/visibility.json": "ff4b672220cc7dd299e4af75d29e521315b120e56e8272d130e922ac05d80164",
+  "source/meta.json": "9ff4688d0bda534fee935e0d5697d0591d3776b27d06d7bc47664034437dc0fe",
+  "source/points_target.npz": "b7d68cb3f88ef455f7b08b89dbd3b16beafec221038365a6bd9fa216189fa2c5",
+  "source/visibility.json": "9cc844372ed5a7f2ed6c308005c18ac9cb30c73da4b807d8abd38898f1092ea9",
   "tracks3d.npz": "699e2802be498d571ac91eaf056f12fa5574ebe9aaa02a204b201cd66fe7b1b8",
   "videos/view0.mp4": "79dd93cd38563a80e12cf49624dc7ade260a9b0610ef6e9dce9b9e885c908354",
   "videos/view1.mp4": "78c250d90b2f0d04fb0ed37f1054e71f8066bda2cce4c2b546c19ce4b5502a47",
diff --git a/Simul-Robot/c001_episode_005100/source/cameras.json b/Simul-Robot/c001_episode_005100/source/cameras.json
index b7942f65f807e766b55e04a2c8fa6fab1cd761c7..f3214ad293656017f316b7f1cfacaa81cffcdee8 100644
--- a/Simul-Robot/c001_episode_005100/source/cameras.json
+++ b/Simul-Robot/c001_episode_005100/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_005100/source/meta.json b/Simul-Robot/c001_episode_005100/source/meta.json
index 433148d56cd6eef96bf0791c2f1b7a16c529c424..9fc4102bf54781a1d5970f67785d56c674cc7804 100644
--- a/Simul-Robot/c001_episode_005100/source/meta.json
+++ b/Simul-Robot/c001_episode_005100/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 1, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 1f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 1, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 1 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_005100/source/visibility.json b/Simul-Robot/c001_episode_005100/source/visibility.json
index 5c76aee0658f337c652af9f8b0274b770699ae7f..16f82be63a08b42ff11d543f7eac163cc5812675 100644
--- a/Simul-Robot/c001_episode_005100/source/visibility.json
+++ b/Simul-Robot/c001_episode_005100/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "alphabet_soup_can": [
@@ -16993,9 +16993,9 @@
       "target": "tomato_soup_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 1, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 1f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 1, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 1 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -17005,12 +17005,12 @@
     },
     "relay": null,
     "hard_target": "tomato_soup_can",
-    "hard_target_note": "타겟 2개 중 primary_object 로 판정",
+    "hard_target_note": "decided on primary_object (of 2 targets)",
     "relabeled": {
       "at": "2026-09-17 14:01:32",
       "was": null,
       "rule": "hard_label2.label(primary=meta.objects.primary_object)",
-      "note": "타겟이 2개라 최초 생산 시 판정을 보류(null)했던 클립이다. 판정 규칙 도입 후 나머지 클립과 같은 자로 재판정했다."
+      "note": "This clip has 2 targets, so its difficulty was left undecided (null) at first production. Once the decision rule was introduced, it was relabeled by the same standard as the other clips."
     }
   },
   "target_world_cm": {
@@ -29715,5 +29715,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_005250/sample.json b/Simul-Robot/c001_episode_005250/sample.json
index 1367fde3567b71e14b5c0ee271f7985193165f4f..9b03bc130b1a52c22f6b0c3b77760112d515b15e 100644
--- a/Simul-Robot/c001_episode_005250/sample.json
+++ b/Simul-Robot/c001_episode_005250/sample.json
@@ -220,11 +220,11 @@
   "depth/depth.zip": "4e899675e5811be8247e45a0777c59ebcf8e9198bad8d02b6cdbc3740ba58365",
   "labels/objects.json": "3e2619189173b53c71f9e977209843df71cac302654be06def740d89d35dd1e0",
   "labels/visibility_states.json": "13c813a1ea2403b635153638e948a17c95a2879dde2abbbb36cbc0def2506f67",
-  "source/cameras.json": "91ce028e0dd76b4b159e4ad3bce059e1632fb2f0d648abae91eb0f1c5a3567e4",
+  "source/cameras.json": "66416d5b36f85ea95608cccdee640cedd380d6b4a5b2da43ee798e114d5e57fa",
   "source/caption.txt": "e9f6883ccd166bfbfbccefbda31748cf249880dab2cfcd1baf5b27e04f1447b1",
-  "source/meta.json": "21e59bba27ca2847239d7d0a5bd23287c560ae116dc4c5a86695313653b7b34a",
-  "source/points_target.npz": "e64a075aa0deaa745a901cde28a372c3ea171781afd033e0af5674eae7074bf9",
-  "source/visibility.json": "202988c2ddf5b6615e78fd60165013fa76531a353d021e4f67e38fcd11ff3114",
+  "source/meta.json": "cfce21cdea54a8eb7d54f790d1fb7988c28276a140ed61a0582bbb5cb29bef15",
+  "source/points_target.npz": "3fd1fc6c1b84c4a7fbbc90ce73ea5da807827d112647a21e4807d865ac142901",
+  "source/visibility.json": "eec2b17ee1ff866efda7a4056bb3f5315a1f61c0bc9fe1bed85aea7c47dbcad7",
   "tracks3d.npz": "6195403115cf7c3db58b2d9d40b02101bf6922ed037601d2a963e99d906c0f74",
   "videos/view0.mp4": "5c7fd71e4d071e2e2df210da29d45266819a303b1edaac2511e4ea2837f37f02",
   "videos/view1.mp4": "ba5872b159a69155aa1d05068b161abcbb151ef0b55f838e41e6731642b3d3f9",
diff --git a/Simul-Robot/c001_episode_005250/source/cameras.json b/Simul-Robot/c001_episode_005250/source/cameras.json
index bf72b90cd3aa543f56b103a6bac2a5584471c991..86e726e2342eb2b6a3ecc2ff668c0f036e07e448 100644
--- a/Simul-Robot/c001_episode_005250/source/cameras.json
+++ b/Simul-Robot/c001_episode_005250/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_005250/source/meta.json b/Simul-Robot/c001_episode_005250/source/meta.json
index a5c0dcd4a9f25157a7ada9ee3ba96561f78fc78c..ba2f87938fd9dd30941999baea278442a1af5053 100644
--- a/Simul-Robot/c001_episode_005250/source/meta.json
+++ b/Simul-Robot/c001_episode_005250/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [1, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [1, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01']: [None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_005250/source/visibility.json b/Simul-Robot/c001_episode_005250/source/visibility.json
index 821f3f0650d4ff975048d4a17a1ac8e684bafb1a..a3a5056fee674509e6b73b843481b9f3053820e0 100644
--- a/Simul-Robot/c001_episode_005250/source/visibility.json
+++ b/Simul-Robot/c001_episode_005250/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "magenta_plastic_crate": [
@@ -16876,8 +16876,8 @@
       "target": "red_milk_carton"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [1, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [1, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01']: [None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -16887,12 +16887,12 @@
     },
     "relay": null,
     "hard_target": "red_milk_carton",
-    "hard_target_note": "타겟 2개 중 primary_object 로 판정",
+    "hard_target_note": "decided on primary_object (of 2 targets)",
     "relabeled": {
       "at": "2026-09-17 14:01:40",
       "was": null,
       "rule": "hard_label2.label(primary=meta.objects.primary_object)",
-      "note": "타겟이 2개라 최초 생산 시 판정을 보류(null)했던 클립이다. 판정 규칙 도입 후 나머지 클립과 같은 자로 재판정했다."
+      "note": "This clip has 2 targets, so its difficulty was left undecided (null) at first production. Once the decision rule was introduced, it was relabeled by the same standard as the other clips."
     }
   },
   "target_world_cm": {
@@ -29507,5 +29507,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_005310/sample.json b/Simul-Robot/c001_episode_005310/sample.json
index e74cf3e9248cc331c9c236c5777808994b7a1c74..ce854e562d871a5fb81e6211bcc734df624ab067 100644
--- a/Simul-Robot/c001_episode_005310/sample.json
+++ b/Simul-Robot/c001_episode_005310/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "3ae510d97d3384bf11f4cec76bbefc0f748ea6f31cfd355c89f35e2a64fe5f7f",
   "labels/objects.json": "93b39117c5f84fa02e65c241dc9a4de81296d028f7e41df1eff9a7d9f0d83f4a",
   "labels/visibility_states.json": "071deac50c7ac634bfda02154594adcdf4bb6e3e259ed4ba9aa1e2019f736595",
-  "source/cameras.json": "36d8a39f769ef3093185a6ef72997c6a436b8a113be1f7409e61cef20612d274",
+  "source/cameras.json": "a5e27394485586d81c5095cfc858408745d0934f2540c6097b91a76d26d07f52",
   "source/caption.txt": "b78bcb49ccb89442201c3fa776c15d6b9a666a5714120668691a994c727e746e",
-  "source/meta.json": "614c61f81ba11bdd8a4e3fe3aea32a69277e052f914acfb1569db76a9d802262",
-  "source/points_target.npz": "5ceb49aff34564e63e8665db860cb1941e681e96543d7ef9c7b61c17e1c5c89a",
-  "source/visibility.json": "be255e54bf8885b119c70ae832b914da0acb8ca28f620a41ba9aa596406890dd",
+  "source/meta.json": "22581c91e5c3daa77c38001f3e4216c570fdd5bc783f29916e639a9cb1b4499a",
+  "source/points_target.npz": "37ef7433d8c15cdf47d74e519ffe3d2cd4f2b9e9dea7240c21608fdbf9fe1758",
+  "source/visibility.json": "e4c52b01d9f36962f68f49d75daabc0b7d203f28a135332ca3ead2d2c15e5c6f",
   "tracks3d.npz": "759e5731e50c1d0bd8545a23e500b5cc997648f24b89ba9f5055332e5f61baa7",
   "videos/view0.mp4": "7b954335ff476b2f01f0dfc4c55036248af7750ddf0ae626ff4186298c8af791",
   "videos/view1.mp4": "96eb3b3e83bb2cd5680ee9341ab2a7a0963f2afe5d2d0f8c48c5f4aaefbf36fe",
diff --git a/Simul-Robot/c001_episode_005310/source/cameras.json b/Simul-Robot/c001_episode_005310/source/cameras.json
index 5a5a247a246292af935a09582fc72b055c7f805a..7f93cb488176321d851531fc9d79fd049fc38975 100644
--- a/Simul-Robot/c001_episode_005310/source/cameras.json
+++ b/Simul-Robot/c001_episode_005310/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_005310/source/meta.json b/Simul-Robot/c001_episode_005310/source/meta.json
index a71e32abe74833ffebf7eb3e5d4ce8c46f1004a7..24c30f8e26bcc8236f1cc2759ac81dd9a79b18e7 100644
--- a/Simul-Robot/c001_episode_005310/source/meta.json
+++ b/Simul-Robot/c001_episode_005310/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [12, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 18f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [12, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 18 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_005310/source/visibility.json b/Simul-Robot/c001_episode_005310/source/visibility.json
index 3e37832e86aabd5a9594f65302f38e75dd5c6245..e79499d026fd472ae6642e1b234b144db796ef4a 100644
--- a/Simul-Robot/c001_episode_005310/source/visibility.json
+++ b/Simul-Robot/c001_episode_005310/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "alphabet_soup_can": [
@@ -16274,9 +16274,9 @@
       "target": "tomato_soup_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [12, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 18f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [12, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 18 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -16286,12 +16286,12 @@
     },
     "relay": null,
     "hard_target": "tomato_soup_can",
-    "hard_target_note": "타겟 2개 중 primary_object 로 판정",
+    "hard_target_note": "decided on primary_object (of 2 targets)",
     "relabeled": {
       "at": "2026-09-17 14:01:46",
       "was": null,
       "rule": "hard_label2.label(primary=meta.objects.primary_object)",
-      "note": "타겟이 2개라 최초 생산 시 판정을 보류(null)했던 클립이다. 판정 규칙 도입 후 나머지 클립과 같은 자로 재판정했다."
+      "note": "This clip has 2 targets, so its difficulty was left undecided (null) at first production. Once the decision rule was introduced, it was relabeled by the same standard as the other clips."
     }
   },
   "target_world_cm": {
@@ -28456,5 +28456,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_005490/sample.json b/Simul-Robot/c001_episode_005490/sample.json
index 297a887ab3fa14dc7b5aad8d38abdce274cbc40a..87cf976711269189c47589ef23998ad1a67288f7 100644
--- a/Simul-Robot/c001_episode_005490/sample.json
+++ b/Simul-Robot/c001_episode_005490/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "1f712babb95829ee75d4957a44c2f9ae0a3b56529f189687f7f50f0b8bb8e629",
   "labels/objects.json": "ea6e644357736fd97fc3480a34b2274608598fb63b01bebc02083be2ebe1e00d",
   "labels/visibility_states.json": "0411ae1723b466e61f982a565808ae364eeec628a17a547449bc1d173f4f15bb",
-  "source/cameras.json": "b4a240c340da6169773548b48607202a30bb556a3c17641118a77fe71b9a1de1",
+  "source/cameras.json": "c30e8151bf476cacbe8b421805d61391d2ecf33b766e53c1f39a4b1da973e562",
   "source/caption.txt": "92a62e710416b00b00ed4f9588ca3303a0c1cf93faf15fe40e79aa4adc760e1e",
-  "source/meta.json": "d61d289dfef5b1204db4af9b84d9945e3fb3afdef1dc18c9fc2d5aad73cb7774",
-  "source/points_target.npz": "cadf3379141e6f5383fbacd81063656621e7aadc888f7785f8914847b26a2fcc",
-  "source/visibility.json": "aa6449019096a968cd2db67a281411946febf1c318e4c187c27c012524d31d2d",
+  "source/meta.json": "ac77311046375faaebeb4db1e5f99d87db6a9c840816417138337937e2c3b0a0",
+  "source/points_target.npz": "96e1c8bccd04e2943a13ef2777bbe7a5594e9247ff7fe163e3d6b346736434e3",
+  "source/visibility.json": "9b868c8dc34e9f77db78c14a18d586796d1d8ffe7a5199c4945596b7c9c90d3d",
   "tracks3d.npz": "a280c8c4fe5a1f9f3a5b7df00ddc957ffcbc4124b0d907177df3c36908bb29e6",
   "videos/view0.mp4": "561a58af60b7df18b0f989f379242db8ee6211e98054365b8eefeae11bdddad0",
   "videos/view1.mp4": "c468d80145d9df555dd2042acc124c0d6190a3a3ff1a4748fbc6eb25db2d5135",
diff --git a/Simul-Robot/c001_episode_005490/source/cameras.json b/Simul-Robot/c001_episode_005490/source/cameras.json
index ef509fd82b581ef2b7e6383d9a5a062715b3b451..4466a9c4c169475f601ed3d4315d27d97572648a 100644
--- a/Simul-Robot/c001_episode_005490/source/cameras.json
+++ b/Simul-Robot/c001_episode_005490/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_005490/source/meta.json b/Simul-Robot/c001_episode_005490/source/meta.json
index 1fbb809f670f53e8b1ac840b01b705586c104621..5d89e402f03650e2d3c011be818ba51a28d37c03 100644
--- a/Simul-Robot/c001_episode_005490/source/meta.json
+++ b/Simul-Robot/c001_episode_005490/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [8, 0, 29]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [3.4, None, 0.16] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [8, 0, 29] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [3.4, None, 0.16] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_005490/source/visibility.json b/Simul-Robot/c001_episode_005490/source/visibility.json
index 1b4af8013a3634d43ee14badaf6c18ee061f1edb..e2f440c9281cc5506fcdd9bb2c9d8f089f083ae7 100644
--- a/Simul-Robot/c001_episode_005490/source/visibility.json
+++ b/Simul-Robot/c001_episode_005490/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "alphabet_soup_can": [
@@ -19117,8 +19117,8 @@
       "target": "alphabet_soup_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [8, 0, 29]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [3.4, None, 0.16] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [8, 0, 29] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [3.4, None, 0.16] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -19128,12 +19128,12 @@
     },
     "relay": null,
     "hard_target": "alphabet_soup_can",
-    "hard_target_note": "타겟 2개 중 primary_object 로 판정",
+    "hard_target_note": "decided on primary_object (of 2 targets)",
     "relabeled": {
       "at": "2026-09-17 14:01:59",
       "was": null,
       "rule": "hard_label2.label(primary=meta.objects.primary_object)",
-      "note": "타겟이 2개라 최초 생산 시 판정을 보류(null)했던 클립이다. 판정 규칙 도입 후 나머지 클립과 같은 자로 재판정했다."
+      "note": "This clip has 2 targets, so its difficulty was left undecided (null) at first production. Once the decision rule was introduced, it was relabeled by the same standard as the other clips."
     }
   },
   "target_world_cm": {
@@ -33428,5 +33428,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_005510/sample.json b/Simul-Robot/c001_episode_005510/sample.json
index 63488c9de3e9c49898a0c532858727c879a16b7b..bcf609f20f873def7ff05f8b8c1c17e99a8c79ec 100644
--- a/Simul-Robot/c001_episode_005510/sample.json
+++ b/Simul-Robot/c001_episode_005510/sample.json
@@ -218,11 +218,11 @@
   "depth/depth.zip": "bd0e67aec4e3d02a0422fee7fb6da927ed44dd9d5b2bde2556662707726e19a6",
   "labels/objects.json": "4d32d92e2733b526851c9fc3d291ef3ec010ca9133c1afd3506e265ad85650fa",
   "labels/visibility_states.json": "faf06d83890d084ae374692376869b50769c1b95f4cf325ac593c47c5eaad9a3",
-  "source/cameras.json": "fe6cbb764f29e64880a0653a1b286db6a2959b32758c07c4dd6e3bd2ab4f6f5f",
+  "source/cameras.json": "4b66e8ebd2989b535606312c818648f8dd11ebf33cd74bf871f116bbd84af27f",
   "source/caption.txt": "77117a6fcbe61e7ff668e05f3b3b8d5d0f01d8fcc180cc1abf319064f38696ae",
-  "source/meta.json": "55e86dc4bf4af7db91cce4b88397ec85b220c44883132eb95b4ca8e11f79edf7",
-  "source/points_target.npz": "3be0eb02c6c2ec9b49bce86ee0db756161d11c6d75f221563441745258963b64",
-  "source/visibility.json": "2878b724981a25f2362900df9590348f144aeb6edea0330cc4c4ac7c0204b91f",
+  "source/meta.json": "07953e795e54c032fa0243adc9fa60f4fad9f75b6f6407531225395bc1fff414",
+  "source/points_target.npz": "6788151380805ce6cfddf07a82c56edde7b97a2903f80992104c9f2d06a09cbe",
+  "source/visibility.json": "a74fadfb7199eb94f3760100b7f2c4f07af9e3db9e718209263dd1be41310465",
   "tracks3d.npz": "80658f0a74f818d2304fb73badd07dd9264d628a8dfc0d76242ad288d9afcc0e",
   "videos/view0.mp4": "862b111f1fc42a92f172841823f5c1801adcae849d31dfcd5eb39324dc09d6f2",
   "videos/view1.mp4": "869579c02ff155c2b9f6d20e1bacfe9db54fe54e2f0be78f4826327875233534",
diff --git a/Simul-Robot/c001_episode_005510/source/cameras.json b/Simul-Robot/c001_episode_005510/source/cameras.json
index 957bb7303434fbbb9207893815a3d29e52fb250a..cd144dfff87bdf4d3f46229f190d30b768627fae 100644
--- a/Simul-Robot/c001_episode_005510/source/cameras.json
+++ b/Simul-Robot/c001_episode_005510/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_005510/source/meta.json b/Simul-Robot/c001_episode_005510/source/meta.json
index aceab2621542a6569fc558ff7371011b5597309e..a1ed3ddeb75168e34d04ae8cf6d7d4509ddfc4cc 100644
--- a/Simul-Robot/c001_episode_005510/source/meta.json
+++ b/Simul-Robot/c001_episode_005510/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [2, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.0, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 2f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [2, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.0, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 2 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_005510/source/visibility.json b/Simul-Robot/c001_episode_005510/source/visibility.json
index d228c39b858c0e880c3e7809a8f0b3f6db988cb4..16d8215ff29d9b6db9dcdea4be622f7b004c78c8 100644
--- a/Simul-Robot/c001_episode_005510/source/visibility.json
+++ b/Simul-Robot/c001_episode_005510/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "granulated_sugar_box": [
@@ -17393,9 +17393,9 @@
       "target": "granulated_sugar_box"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [2, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.0, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 2f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [2, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.0, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 2 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -17405,12 +17405,12 @@
     },
     "relay": null,
     "hard_target": "granulated_sugar_box",
-    "hard_target_note": "타겟 2개 중 primary_object 로 판정",
+    "hard_target_note": "decided on primary_object (of 2 targets)",
     "relabeled": {
       "at": "2026-09-17 14:02:00",
       "was": null,
       "rule": "hard_label2.label(primary=meta.objects.primary_object)",
-      "note": "타겟이 2개라 최초 생산 시 판정을 보류(null)했던 클립이다. 판정 규칙 도입 후 나머지 클립과 같은 자로 재판정했다."
+      "note": "This clip has 2 targets, so its difficulty was left undecided (null) at first production. Once the decision rule was introduced, it was relabeled by the same standard as the other clips."
     }
   },
   "target_world_cm": {
@@ -30415,5 +30415,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_005550/sample.json b/Simul-Robot/c001_episode_005550/sample.json
index 601f4f68a07f43281f10871cec9518655e805ba7..7a63d1d47d4758dec2851659924d6d8da7d96c97 100644
--- a/Simul-Robot/c001_episode_005550/sample.json
+++ b/Simul-Robot/c001_episode_005550/sample.json
@@ -218,11 +218,11 @@
   "depth/depth.zip": "de9e5a37bf97e9fcca4d507a1d2928f797a41f536d173a53ead6353311a1a216",
   "labels/objects.json": "88cb6493733211c0e8ab7e073d08da3bb11e413f9fc39b5c2f1663bdf716f306",
   "labels/visibility_states.json": "96a6cc06b3a63a44158e8b1d8344718e265a7cf71ae637b7363ef912a5f75700",
-  "source/cameras.json": "c2b2445e5abf130ffefd15ca2a1f9074d3122416ee3593b00e68dd47ad1db0d0",
+  "source/cameras.json": "95b37760071887007ed9aa684040d0f1d1b0a30579d0c5d7015856d33920df0a",
   "source/caption.txt": "265bf8c3ae557111bce8ce26321be57873306eb369c78db32f006d1737a297ea",
-  "source/meta.json": "4f24beeb1b13d9dbef1cb12a95599c43549668759d70c950cedd51b8b49f0bcf",
-  "source/points_target.npz": "f79f164f92e927d7ae6a7dc6e4ee44c8135d9e67ce96dcfd77dfa6f2c49e3e33",
-  "source/visibility.json": "384bd91f81e18026a0d840a99a55318760484bd326d8b5aa549a406eaad94836",
+  "source/meta.json": "d0b281a87ea6c341d94163d249dbbd8607bd3768028f541537e65430df3c5073",
+  "source/points_target.npz": "76c959ce0c8a77c42c0de87a84c0200786ead34b8bcd0ed902a19a920cd6baf8",
+  "source/visibility.json": "aacfe6c8ba5e185612f4014195a5c255dafef4543ccbd1c2283c81f81bc5c438",
   "tracks3d.npz": "e55077baf374c3aad47358fbf98a4d4e842a0311166eb219c58c46acef41141b",
   "videos/view0.mp4": "6d06e47aee7e22a21c6cca50a6f82f78dab10148f473ed4b81fe740941914590",
   "videos/view1.mp4": "2fa8d21e2c1620bac5bf616458715d47a0cfba8e2aa8cc804e5b884b3565699a",
diff --git a/Simul-Robot/c001_episode_005550/source/cameras.json b/Simul-Robot/c001_episode_005550/source/cameras.json
index f4dc530c51311e0dfc0b64e56aa6b595fa1b890c..d47ecc076e877af18afc9f4f742aae833c14530c 100644
--- a/Simul-Robot/c001_episode_005550/source/cameras.json
+++ b/Simul-Robot/c001_episode_005550/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_005550/source/meta.json b/Simul-Robot/c001_episode_005550/source/meta.json
index 9b1e0bb5d20d91c5df3ccf00cb49ee39b238ab51..3edbb8e652b3d8100ee16c5449ab270c3a3b37e4 100644
--- a/Simul-Robot/c001_episode_005550/source/meta.json
+++ b/Simul-Robot/c001_episode_005550/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 2,
     "hard_reasons": [
-      "①캠 ['cam00'] 의 최장 비가시 [20]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, 0.96] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00']: [20] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [0.01, 0.96] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_005550/source/visibility.json b/Simul-Robot/c001_episode_005550/source/visibility.json
index bcde18600e367176ee640924733e57786976e38f..5e3a224d4af35b1aed02aeee92340a620e4bb5d2 100644
--- a/Simul-Robot/c001_episode_005550/source/visibility.json
+++ b/Simul-Robot/c001_episode_005550/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "dark_foam_brick": [
@@ -17319,8 +17319,8 @@
       "target": "dark_foam_brick"
     },
     "hard_reasons": [
-      "①캠 ['cam00'] 의 최장 비가시 [20]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, 0.96] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00']: [20] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [0.01, 0.96] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -17330,12 +17330,12 @@
     },
     "relay": null,
     "hard_target": "dark_foam_brick",
-    "hard_target_note": "타겟 2개 중 primary_object 로 판정",
+    "hard_target_note": "decided on primary_object (of 2 targets)",
     "relabeled": {
       "at": "2026-09-17 14:02:04",
       "was": null,
       "rule": "hard_label2.label(primary=meta.objects.primary_object)",
-      "note": "타겟이 2개라 최초 생산 시 판정을 보류(null)했던 클립이다. 판정 규칙 도입 후 나머지 클립과 같은 자로 재판정했다."
+      "note": "This clip has 2 targets, so its difficulty was left undecided (null) at first production. Once the decision rule was introduced, it was relabeled by the same standard as the other clips."
     }
   },
   "target_world_cm": {
@@ -30280,5 +30280,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_005580/sample.json b/Simul-Robot/c001_episode_005580/sample.json
index 8493de46bc3c4afa7bad296db1d15b0abd371c23..34a0ab57bead9e4e6c31003ba87188663bcec6f4 100644
--- a/Simul-Robot/c001_episode_005580/sample.json
+++ b/Simul-Robot/c001_episode_005580/sample.json
@@ -218,11 +218,11 @@
   "depth/depth.zip": "37c29281c037d93825a58a782eb4b873e6730b1127d00f4a425f86d6b082f9a6",
   "labels/objects.json": "37b475125f4972f4719397717c86aeeb7ec5452beeef9b99fa2a8be5f70598a7",
   "labels/visibility_states.json": "42b5f286115dfc919eb3ea10a024ffca4ce95d99080017bc6bb9309fcf926544",
-  "source/cameras.json": "96e6da49348ebe9a140a9ecb45eefb36f9d1aab20c69c98075ba2aa361ea071c",
+  "source/cameras.json": "0b7eb4d90aac15e77c08a111fc0c813d4cbe43e43b64936af463783d5e5d1b2b",
   "source/caption.txt": "0abc9d3d586c54f2c447fdda16148c056a76f60320e338d09ce62a09d06eb3fa",
-  "source/meta.json": "86f18a4ae19522b737181c07e87abc711a504ef42677be3f8c291ef44312a7c7",
-  "source/points_target.npz": "2bba14bc5208ae3fe1aac950f66d554e6d1e470b59cb28a76c36d03c17ad1704",
-  "source/visibility.json": "6c7832f846d20641b74f14ffa85f502c55c20640addad81e86f8f640c17d0331",
+  "source/meta.json": "b3479b00c056be5c59acd69c4bf129471b952ae2054dc5600ed07734c8913919",
+  "source/points_target.npz": "e6e71157203459088047c5ae2dd8d2490736aac8111c056d8d45cc972f1eba00",
+  "source/visibility.json": "54e414330271dadd28cb911c3f1e4f3d9333bc16b9794f2dd53302c8b28f57e3",
   "tracks3d.npz": "4b30fb7aef0fb6755ba85b2c1714308690164acaf157c66b13c27ff8124f7911",
   "videos/view0.mp4": "bf4450de19c94344f8ded4ea844afc5c5ef4e76a9af027c6419ca0f45afdf34c",
   "videos/view1.mp4": "ebab895b5e68346417ab106d8307166f2ef2feda03f7477a8a7e74d124926511",
diff --git a/Simul-Robot/c001_episode_005580/source/cameras.json b/Simul-Robot/c001_episode_005580/source/cameras.json
index 52b0ec487f9073fb0dc7ebecda4a527f07bf32de..24263e8b38f8c81935c2ed58dc7648646fbef442 100644
--- a/Simul-Robot/c001_episode_005580/source/cameras.json
+++ b/Simul-Robot/c001_episode_005580/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_005580/source/meta.json b/Simul-Robot/c001_episode_005580/source/meta.json
index ab4b56fc83166526e8e54c0022d9c4baab212e1c..a87cd18592b8a22bede8a7c2c3c29ed648de7a60 100644
--- a/Simul-Robot/c001_episode_005580/source/meta.json
+++ b/Simul-Robot/c001_episode_005580/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [4, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 5f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [4, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 5 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_005580/source/visibility.json b/Simul-Robot/c001_episode_005580/source/visibility.json
index 805a37ea9963305fdd434bfb028d4e17d5bd68c1..cd4a10ed36342c7b835fda42f1980d82524d5987 100644
--- a/Simul-Robot/c001_episode_005580/source/visibility.json
+++ b/Simul-Robot/c001_episode_005580/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "green_tuna_can": [
@@ -16193,9 +16193,9 @@
       "target": "potted_meat_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [4, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 5f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [4, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 5 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -28307,5 +28307,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_005680/sample.json b/Simul-Robot/c001_episode_005680/sample.json
index 271f695a51e65199dd1d05e5bd5e7a15196d1a6a..0152ab82479fae828a0d8bfd3e9d6ae510fa3062 100644
--- a/Simul-Robot/c001_episode_005680/sample.json
+++ b/Simul-Robot/c001_episode_005680/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "796b0ff2b7cc231836da31e875f867a8c22551c72388dacfa9a39f9a8e72cc12",
   "labels/objects.json": "f7a95c6361e1c998d1ebf89e8f110a3ac55c17aaabd0cfc31ac83ffb44b8dcb9",
   "labels/visibility_states.json": "92ace864b8eb334b2794ef14965a5aa39f073dfcc52d12ac3e47af8f4fc63317",
-  "source/cameras.json": "c1adb76dc931a6789dcc88784d22cbd4abc1184d5d03d5c406ad04b0fd0082ae",
+  "source/cameras.json": "a1afbddaee257e7e15685feecd4a0ad6d6a1fe26fd644129c8d358c22c2792a2",
   "source/caption.txt": "5e1fae365209ab273336041caed779544a8aeee591ebc9a2776f6299de589678",
-  "source/meta.json": "00fa6fd8dbeff1fbf3a321023a858eb932af6d0557490610b52663981df1d1dd",
-  "source/points_target.npz": "ce7b944a3976205610562235c99857213f7e0f3a462fedcb298d550ca9976f14",
-  "source/visibility.json": "90f9a40aca88df16eb62ba0b92da35523cfa1de4d4bd78df3b3e37cda72cce98",
+  "source/meta.json": "f77a2c14367b97d422e39813b402ac69db60f6c62ab8a558f73bb6f221c334c5",
+  "source/points_target.npz": "7dc496ff6efa075e37d7fc41f68305ce9bedbec00afa523678b9238be3cfd852",
+  "source/visibility.json": "9c9c6650bbd24f16de4893bcbe98b437b19c6a070d67e6ec81302ccf83529cf9",
   "tracks3d.npz": "bc6ec790c713bdc0444ce052cb8e971cfb682c3194047001ebe196e6794810d7",
   "videos/view0.mp4": "e0233c33d259e1d7c785d7ce12ef84f4ead6b075020f3b8e8afff2a0f4f8a472",
   "videos/view1.mp4": "21a9e066a45cca0b56ec34ecb8eed91abaad4127fb5cf8aa86d3a031fcd36313",
diff --git a/Simul-Robot/c001_episode_005680/source/cameras.json b/Simul-Robot/c001_episode_005680/source/cameras.json
index 9cb0d30eced38c45775d1b5ec4adf094550cc910..506d339ecff0afd637442b16ebb53cd81909597c 100644
--- a/Simul-Robot/c001_episode_005680/source/cameras.json
+++ b/Simul-Robot/c001_episode_005680/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_005680/source/meta.json b/Simul-Robot/c001_episode_005680/source/meta.json
index 74d68ceeb1fbd9f9c422a1f5722f2b058fbae4f5..3e241257ecb2c5dfdd3d2d143e5ed19a18f4ee32 100644
--- a/Simul-Robot/c001_episode_005680/source/meta.json
+++ b/Simul-Robot/c001_episode_005680/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_005680/source/visibility.json b/Simul-Robot/c001_episode_005680/source/visibility.json
index 8041f1340b8c6a7868e7e22c19b0027198c165db..9eec4f764a86204424af266d6a3b3d3165b59920 100644
--- a/Simul-Robot/c001_episode_005680/source/visibility.json
+++ b/Simul-Robot/c001_episode_005680/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "alphabet_soup_can": [
@@ -17191,9 +17191,9 @@
       "target": "tomato_soup_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -30055,5 +30055,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_005930/sample.json b/Simul-Robot/c001_episode_005930/sample.json
index 578dcd93f2aec169be9a3832292a6f7402ec60ab..07bd7df882f780f6015dd5a6005b8335111e3fbd 100644
--- a/Simul-Robot/c001_episode_005930/sample.json
+++ b/Simul-Robot/c001_episode_005930/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "14e9e934787b42c920a96044b05a171d0f1120feeb5424828676283f99779a7b",
   "labels/objects.json": "c5f925b9cc53fd1bf96f45f4bc670e4989e708ba149e97ad2ad785f252c53873",
   "labels/visibility_states.json": "0b5840e0afe21fef77fb1161644bf48e76827ee0b6897461cfec18258010f30a",
-  "source/cameras.json": "3c6e71da175c0a50a21a9c1522a571f08a8d6b31102b007707f2a6548fcfb7a4",
+  "source/cameras.json": "0dd6a5169304a10b64ad300e841eeece77c2fecaa6e4780a17374c62c68a99dd",
   "source/caption.txt": "b5e587a5a7a1d50b454462e1af7e1b1df24474ee8e49f5b8d8aa2494d7dac389",
-  "source/meta.json": "24feb2d17786c25ee4395fa4b5342783a83ea81df288d2dc8529c7c839f8b0ad",
-  "source/points_target.npz": "d308bc65033c45e00b2ec25ec26721f83a9490eea873d7be230c8db11b70c9a5",
-  "source/visibility.json": "93e1bad57cef582a8f4fbe430d5d95ed9379537e2477ebd393166dee93cc5d77",
+  "source/meta.json": "de4afcd03d21d399a033ef6861d366e356b17eba7ba15cb0b4a7134c052abe0f",
+  "source/points_target.npz": "3d35272ed2e781646a758eaff46a2f10b6d811a9e05d9b705cd9d7e1729bad0c",
+  "source/visibility.json": "4fed432e9405f95ff956888bbc5fa4a74547fa45d65df23688dc58268d83d372",
   "tracks3d.npz": "fde86ddba8468b0e3e926fb308780c608a5aae106d989edf16f868424f46e6fb",
   "videos/view0.mp4": "baa14ed3a70da09aa68cd657134f68013f02035b1c7ad1336da2fa0e438f9fb3",
   "videos/view1.mp4": "b45a50823edde8d3e0516dd6dc16ed42f03ddca2f731cb4a9cc726d316ef0b9e",
diff --git a/Simul-Robot/c001_episode_005930/source/cameras.json b/Simul-Robot/c001_episode_005930/source/cameras.json
index 9f1249e64890a82f7c8574c1df5f3a5723a695fc..cc08ac17dc798616fd317f596b1a9520fbbf2f5f 100644
--- a/Simul-Robot/c001_episode_005930/source/cameras.json
+++ b/Simul-Robot/c001_episode_005930/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_005930/source/meta.json b/Simul-Robot/c001_episode_005930/source/meta.json
index ebb60b4490c081879b43c7fa70df4825bf9cdb6a..6301e0b217b1a9ee849e99e98008fcfb55be9fdb 100644
--- a/Simul-Robot/c001_episode_005930/source/meta.json
+++ b/Simul-Robot/c001_episode_005930/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [22, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [22, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_005930/source/visibility.json b/Simul-Robot/c001_episode_005930/source/visibility.json
index 4ca50b0f3a8784a0a671293f1dc1824465f64403..63ea247552918544228d20cecfb425feeb4fc23b 100644
--- a/Simul-Robot/c001_episode_005930/source/visibility.json
+++ b/Simul-Robot/c001_episode_005930/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "dark_foam_brick": [
@@ -18354,8 +18354,8 @@
       "target": "dark_foam_brick"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [22, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [22, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -32087,5 +32087,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006130/sample.json b/Simul-Robot/c001_episode_006130/sample.json
index 94af9db7a8d10372a000715cf01a48e55e6e6950..1f96df4b85666d1b17d0eac6d0a91a56371ebc29 100644
--- a/Simul-Robot/c001_episode_006130/sample.json
+++ b/Simul-Robot/c001_episode_006130/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "75bf1c0eb12c6dc97bd95192f89544852e42486672662b789814ece3420320fe",
   "labels/objects.json": "bc3e1526616e196abf33ac268b307edec5b63af7fabb456ceb48f6ba85d0627f",
   "labels/visibility_states.json": "fcb1860464f9a78b1ef731732b833aa93593d55a3a22adb948d46e774fb28bf0",
-  "source/cameras.json": "35b8564c6037fef6ea7c2ec90f832100c8d83864aca6bee387b2f98366173f20",
+  "source/cameras.json": "bea5b034bbe329ac486753f121633ae031bc411c483f2adb670bbdcb2cc33f35",
   "source/caption.txt": "f2b0ed719b9fbe834e58daaaf29d5efcbd1eee3c1820367bcf90e88d533b32f7",
-  "source/meta.json": "1b46b8ef3a3e464797600b85c337bcb95bf69bd6ef1347c9b520f971ea193643",
-  "source/points_target.npz": "b062841c8d1ee2456c6c49f12ff77b35cc437e5d5934cb3d01d4c3715d9fce34",
-  "source/visibility.json": "0d548e5dcb8fb52a60cb970d7e17f3f284998fca1bd8fd2dc4aa03b2bbd62cc2",
+  "source/meta.json": "2ed2b0979bd6ba1052265492df07609092ff83f79f4ba8ad1637bc18fbacdf22",
+  "source/points_target.npz": "af05bf25b3ecf8bbb9e441a32da732dfd74efcec1c66ba290e3d9d47de43c21f",
+  "source/visibility.json": "efb4c89efb2411108ea2178768f707707a318e7b0689ba673b8c0b3570f16fc6",
   "tracks3d.npz": "de9c26b9542438b1fc218a270765702d104a11fe24eef0d76887e539214b9175",
   "videos/view0.mp4": "ea7f8a8c1bb0813622e61c995f852d5a6091e15c64672a92e5d8bb56c9f19d7e",
   "videos/view1.mp4": "a37924acee82bf295784d84752435ac8008babd514bd5ae55ade84e40461bc4f",
diff --git a/Simul-Robot/c001_episode_006130/source/cameras.json b/Simul-Robot/c001_episode_006130/source/cameras.json
index 854132daa6dca6c89e0bbe3d9a5857444afba341..e821d34ade6e41849e689893983b4c51a38493de 100644
--- a/Simul-Robot/c001_episode_006130/source/cameras.json
+++ b/Simul-Robot/c001_episode_006130/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006130/source/meta.json b/Simul-Robot/c001_episode_006130/source/meta.json
index e250a2ab949a8296de639ffd4290ea4cab2060d9..c67216695125480b76445f7030598718ac9fc3b8 100644
--- a/Simul-Robot/c001_episode_006130/source/meta.json
+++ b/Simul-Robot/c001_episode_006130/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 2,
     "hard_reasons": [
-      "①캠 ['cam01', 'cam02'] 의 최장 비가시 [16, 23]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01'] 의 비가시 구간 타겟 변위 [0.01] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam01', 'cam02']: [16, 23] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01']: [0.01] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006130/source/visibility.json b/Simul-Robot/c001_episode_006130/source/visibility.json
index bd4bef11c73cdb26cc3d9eaee9d508c32bc1fb8e..1400aebedb007129e64e3bdb1fcf5dee8ba4328f 100644
--- a/Simul-Robot/c001_episode_006130/source/visibility.json
+++ b/Simul-Robot/c001_episode_006130/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "orange_juice_carton": [
@@ -17000,8 +17000,8 @@
       "target": "peas_food_can"
     },
     "hard_reasons": [
-      "①캠 ['cam01', 'cam02'] 의 최장 비가시 [16, 23]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01'] 의 비가시 구간 타겟 변위 [0.01] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam01', 'cam02']: [16, 23] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01']: [0.01] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -29713,5 +29713,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006210/sample.json b/Simul-Robot/c001_episode_006210/sample.json
index 04cfa9d16217ba90cc07bedb347fb0f85a0d4a4f..9c2698cc7790a57433f70e255958a5fa74ad3492 100644
--- a/Simul-Robot/c001_episode_006210/sample.json
+++ b/Simul-Robot/c001_episode_006210/sample.json
@@ -218,11 +218,11 @@
   "depth/depth.zip": "ed11a80a8b2334d5fd5f8da824abcde57199cd5bd45a51de5dcacc5a038fceb8",
   "labels/objects.json": "3ee8413f43e7726212b54e6272dfadd7e7f6dd4853a03cf7d14262c4410c9cf6",
   "labels/visibility_states.json": "4ef633c4356c8431c944dcba012aac264b7efe8060f2f7f121b9260b604b1594",
-  "source/cameras.json": "3e9e8407a5a9dbb48c322cb3aa829c290a9dca694b72529a9681f8947688d131",
+  "source/cameras.json": "6c0e3b63add26e3afa95ff3d821e9bc5ae77f0b02cd7577c20ea73f842367f41",
   "source/caption.txt": "c4bfef9c056969ada9bd038ca002127e6db4fb81047eba70ef2c563392436f6b",
-  "source/meta.json": "987f4c52255d7f8f69faa9b328cc8ac489dc2aceed11636a03330ce9d53d9bf8",
-  "source/points_target.npz": "c7eb331dbcd5a8831c72bb1c9ee5c9484082834206b194eea80e94df679f6637",
-  "source/visibility.json": "0a85cda27085435b41fd20472038860f1782f7c1f547e79440194c938c94f432",
+  "source/meta.json": "0745d25a744fd4cf3adeeae615f2d6ac9f2a9ad9e739ad912d4ccce1868b67ec",
+  "source/points_target.npz": "29a4940d3df8fcddac60bef07fecdf7067dd24a134215ce28c714fda374706d9",
+  "source/visibility.json": "a2b18f49e3a85e2d85c64cb06ebf339ffac8dfe6cd0d92a8a806679fe61f46e3",
   "tracks3d.npz": "3fec293c6e0a86d617fbf70f7196112c721c4f8d6a04d2d9e2a0a1f6577f4982",
   "videos/view0.mp4": "59ed737b6f0599f346c6256d127911544384e669b55a7d68c471398f712df324",
   "videos/view1.mp4": "0da0c014d39d2907907b06210355fcd9521c7b36434240fa0224dd2ed8a5aab1",
diff --git a/Simul-Robot/c001_episode_006210/source/cameras.json b/Simul-Robot/c001_episode_006210/source/cameras.json
index 9e1f2e7a946a2ba87c341aa6c1932b49d982d2f6..47e3be3cc7b86ba2b55e085159b57c4c40f4198b 100644
--- a/Simul-Robot/c001_episode_006210/source/cameras.json
+++ b/Simul-Robot/c001_episode_006210/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006210/source/meta.json b/Simul-Robot/c001_episode_006210/source/meta.json
index 5cb07c964f7c92794ae716e64b5dcf768397467c..7359db679bf29a9f5ab4720c9a64647b7bd57b1b 100644
--- a/Simul-Robot/c001_episode_006210/source/meta.json
+++ b/Simul-Robot/c001_episode_006210/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006210/source/visibility.json b/Simul-Robot/c001_episode_006210/source/visibility.json
index 147ad4746f624aab878b36676cc407510dacb9e1..3d4e14562881130ab72b7a79c6f10eba0a2f527b 100644
--- a/Simul-Robot/c001_episode_006210/source/visibility.json
+++ b/Simul-Robot/c001_episode_006210/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_tuna_can": [
@@ -14950,9 +14950,9 @@
       "target": "white_cat_figurine"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -26274,5 +26274,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006280/sample.json b/Simul-Robot/c001_episode_006280/sample.json
index 55b6058d0b94a231ec06f8edf6d44a2e23597763..1522d08d800e801dc63612ff2edae372b3112610 100644
--- a/Simul-Robot/c001_episode_006280/sample.json
+++ b/Simul-Robot/c001_episode_006280/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "afae73c212ef8ad3fa0e8642ab899432738669b61ec34ee8462cc10f776ead9d",
   "labels/objects.json": "6d2fcb4da94334e0fe98bb7b83cf5c8ed2e1b01702751a66a6cc419b7d06c442",
   "labels/visibility_states.json": "565919aaab6985da8f67fad711f2a209079a1fff6b13de36045cd4208f372e1d",
-  "source/cameras.json": "8289fbd122601e93bb6dd9afeb2aa3384a49defab40f6120a26f71d72f6bdc56",
+  "source/cameras.json": "500f914e0c47dfaeed0b9eebe679c731df263ffe4bf49b730a552c4932dc22a0",
   "source/caption.txt": "c539ae507da1fc0572447cc0e946b99a8caa9d7753e037adfa2882dac98a4c08",
-  "source/meta.json": "a4ccda2174cefdc245d11440726c250dd760681345c14addab955ec70d6cadbf",
-  "source/points_target.npz": "cf444afbd9ca50226a5ae3ba95fca4a53c44bddfd6fd0f99baf3b5c3aa0758d3",
-  "source/visibility.json": "57e9ad167a529de5775b558467a93505979f5144b62adc26ebb1ce9ebfce31fa",
+  "source/meta.json": "888473af4848b2938b12d7e5b5c9d9c2191785ded205532b5001b5f61dca7e11",
+  "source/points_target.npz": "1bccdcdda011f81ad41122aacc6797a203b816816cdb5e94e794c69e34a230fc",
+  "source/visibility.json": "2b96a347505ac1f2e587ba3e17987c6b6cf0fecca5404a679828e629b629370e",
   "tracks3d.npz": "9126dfefccf9efabc0185827d35f9dc3d6c1d63a06969b21c930ac9e9bfb6a69",
   "videos/view0.mp4": "436d0f40ab03401156f2ad06c88e4360ee251336308ceb44b3d2903a223e3cdf",
   "videos/view1.mp4": "b33c6284208653cc1ba23b4ea2940b898d3003279d9f202be86a1e34cadba0ab",
diff --git a/Simul-Robot/c001_episode_006280/source/cameras.json b/Simul-Robot/c001_episode_006280/source/cameras.json
index ec0242b98b07a19e637dc30c63a08730bb9e218f..7752c71b11af80c9363192a81d6386b373effb84 100644
--- a/Simul-Robot/c001_episode_006280/source/cameras.json
+++ b/Simul-Robot/c001_episode_006280/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006280/source/meta.json b/Simul-Robot/c001_episode_006280/source/meta.json
index 87f8d8e334fcb244b8131d1d02060fee60a531a2..6ef745d7244ae3fc1d1fc4019a193e3cb7816521 100644
--- a/Simul-Robot/c001_episode_006280/source/meta.json
+++ b/Simul-Robot/c001_episode_006280/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 2,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [17, 3, 7]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [3.02, 0.54, 1.12] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [17, 3, 7] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [3.02, 0.54, 1.12] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006280/source/visibility.json b/Simul-Robot/c001_episode_006280/source/visibility.json
index e05a5e2cdcc6309b730238695f65bee04a3431fb..d19631fc8617ea367debd3f9efeeca9f2e61b4e6 100644
--- a/Simul-Robot/c001_episode_006280/source/visibility.json
+++ b/Simul-Robot/c001_episode_006280/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "magenta_plastic_crate": [
@@ -17760,8 +17760,8 @@
       "target": "parmesan_cheese_canister"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [17, 3, 7]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [3.02, 0.54, 1.12] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [17, 3, 7] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [3.02, 0.54, 1.12] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -31043,5 +31043,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006360/sample.json b/Simul-Robot/c001_episode_006360/sample.json
index f73c0102af1f0e0e5fc14322ff97bee85c837912..c8b2640ec165c7e155608bb7729f4e9ca8bd6d8b 100644
--- a/Simul-Robot/c001_episode_006360/sample.json
+++ b/Simul-Robot/c001_episode_006360/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "b1e33b404fb922914df07619e24cf68a8a340415f43a0820341dd1210766bf3c",
   "labels/objects.json": "b18a636d039f17d4d85a331a037770e6fa5d826b77031f019837796d4acab1d0",
   "labels/visibility_states.json": "2fcd9b028aac359070638afe488ad929a729d386d65e297a1fa2aaa3d2411eb8",
-  "source/cameras.json": "cf020d0046da1f6435117ee513d8365c8075f7ec741f1a1566892fed1fcf73fa",
+  "source/cameras.json": "34c2491d3f9d02e26f5c4f6ebbcf48ac8d6fd494f5b3e114346fc574169c8d97",
   "source/caption.txt": "20d3597e6ae69b42871e29893e29964c985a93883c024c755d10d4c85e9784dc",
-  "source/meta.json": "88c9e47193d8766c381bce60048ffd86f8be07c4d481e9af2c4ba037626c7162",
-  "source/points_target.npz": "db9866d849ddc293c5e66f3b3f43ce3bffffc46060a4385b465e403d0d49b2c6",
-  "source/visibility.json": "fce59855a15ae751cbf6bfeebf876e116e8b7e051d473729e33c83ebbd6efe00",
+  "source/meta.json": "a23a3aac0890ed7a3de6f98bdc75e265acd4bb13db77d8ea91178ff2105d6043",
+  "source/points_target.npz": "c36be513241fd77e4f42a24897491597c4dbfd269e717931806584419c332515",
+  "source/visibility.json": "60cb857abc7546885dbe0af9ef790a4808cd4a8ed723500e4544f7eaa83283a0",
   "tracks3d.npz": "0e02c85a3398149c55bad625303c2c286296ca3095ebd36b3531997b2f9e5eec",
   "videos/view0.mp4": "10ed44c1b25b9dda203ca561b18c5c660e65af632da24dc92c303c0d63a08613",
   "videos/view1.mp4": "f998587ad12f56d793d8df8c68b4178920b92eac97829d3e6748cfee806d38f9",
diff --git a/Simul-Robot/c001_episode_006360/source/cameras.json b/Simul-Robot/c001_episode_006360/source/cameras.json
index 9d3640aeba48096405ee86ca07c3fc2a393c7ad6..51183db6772103fdab2061bf5fa033b8cb7f0f98 100644
--- a/Simul-Robot/c001_episode_006360/source/cameras.json
+++ b/Simul-Robot/c001_episode_006360/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006360/source/meta.json b/Simul-Robot/c001_episode_006360/source/meta.json
index e785d8d6744fc19a2b28ee2c5619fcaec0dd0edf..e6979b8862a9f596ee7995b4728a40a666c1498e 100644
--- a/Simul-Robot/c001_episode_006360/source/meta.json
+++ b/Simul-Robot/c001_episode_006360/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [17, 1, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [17, 1, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006360/source/visibility.json b/Simul-Robot/c001_episode_006360/source/visibility.json
index 94212f02620bffb5f3dac451e3dccd6b76dcdd62..8af8353b5fb0e77333dc36ac4a20ae3309b9b7ca 100644
--- a/Simul-Robot/c001_episode_006360/source/visibility.json
+++ b/Simul-Robot/c001_episode_006360/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "art_deco_bust": [
@@ -16157,8 +16157,8 @@
       "target": "potted_meat_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [17, 1, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [17, 1, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -28240,5 +28240,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006370/sample.json b/Simul-Robot/c001_episode_006370/sample.json
index a03ff77e4bf36cdd6fdc6402afdaeb9a6e071e58..932818945eaaec31c88e24fb56cf1fc276b5a6cd 100644
--- a/Simul-Robot/c001_episode_006370/sample.json
+++ b/Simul-Robot/c001_episode_006370/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "d03a09f35793d40c37d817c623f9a168c12860406319fa053c461dc04ce1b153",
   "labels/objects.json": "3bab0ed1aa646f1011cd704a3a57f32eeee39ab0b9ca2e09443799e1194f3eb2",
   "labels/visibility_states.json": "c685ea34d4693914c2470e3a48fdb00f3010c2bad9c7b7206df3b8a0e6a6bfa5",
-  "source/cameras.json": "d30c271f93053a5954743442174fb152b31bc20d7b23013eb2a65a9076919e8c",
+  "source/cameras.json": "e474753fbbc101e2ad926cfff888bd74fccb0c82afb72401e90dace85a6f4042",
   "source/caption.txt": "067eef3b907f09bba6cedcf22f1bc10654677777d6ec9abf0e8061ec3a808f60",
-  "source/meta.json": "5cd51c677c03c91526465e6bef46d70a3ea1c1790bc517e04d1922d5619c5545",
-  "source/points_target.npz": "89648340f46a4ea4e14ff06cd6bd7cdc766e0e5dd9506b79786f34402899d4d2",
-  "source/visibility.json": "1407994e5800fabe5693b840c8fb7c72828179bbd538e739ec3fe46941153344",
+  "source/meta.json": "a0561e71dc85b7836fe4296067712f753252cdffead2301af45bbc07070377df",
+  "source/points_target.npz": "c11cccbab3842d7aaa54a4a6b7872abcf30d6a6531a98cc4e9469c7686b2b696",
+  "source/visibility.json": "493c21f04d5498fb1b4e08bea42076f7684d66daa2291ec9ee488ac1da39dd92",
   "tracks3d.npz": "492ad78d63b644ad68cb55cf677348ac9bb2dc69b18839ca4adbdb97d7b894e8",
   "videos/view0.mp4": "0401b61d7bf851d05f7d83f6058369963311967bfb885b2a2a12bc3b38ff21c9",
   "videos/view1.mp4": "0a8d3587aa95ce39eae7a3514bac47b71c7edb32c71d8e367a608f59bbb0f1a9",
diff --git a/Simul-Robot/c001_episode_006370/source/cameras.json b/Simul-Robot/c001_episode_006370/source/cameras.json
index 51fbd0db04e67dbff88d8e11a64351cfa2bed7a5..f87869b5533b0ced3ab98ded4861e5a39dd04f8f 100644
--- a/Simul-Robot/c001_episode_006370/source/cameras.json
+++ b/Simul-Robot/c001_episode_006370/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006370/source/meta.json b/Simul-Robot/c001_episode_006370/source/meta.json
index eda661b1d5a01d6540c36bfcbb233f909a4bc08e..9fe3c7c8f01fa027a1913f758c92d173c97a48d9 100644
--- a/Simul-Robot/c001_episode_006370/source/meta.json
+++ b/Simul-Robot/c001_episode_006370/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [11, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 12f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [11, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 12 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006370/source/visibility.json b/Simul-Robot/c001_episode_006370/source/visibility.json
index 2d89b0267ce077904949ae6ff1acc0e74d126870..0e7c585c26cb3e0fab4e9f9eeea4ce7db0b66d7d 100644
--- a/Simul-Robot/c001_episode_006370/source/visibility.json
+++ b/Simul-Robot/c001_episode_006370/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "dark_foam_brick": [
@@ -20594,9 +20594,9 @@
       "target": "dark_foam_brick"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [11, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 12f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [11, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 12 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -36008,5 +36008,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006390/sample.json b/Simul-Robot/c001_episode_006390/sample.json
index 030c33e91e98294b7e9b0f96ce4874e96b275510..425ebf2c002eec36fc9791bf90f8f1dfe4507675 100644
--- a/Simul-Robot/c001_episode_006390/sample.json
+++ b/Simul-Robot/c001_episode_006390/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "cf5c6eab514101313546711c9ea6bddc2c2c524a16ca1a05e5cf79bfd7a02cc9",
   "labels/objects.json": "077f40e496ce0f97dd3caca2228b92b3da40bc61d0f1811708496b9355475482",
   "labels/visibility_states.json": "e0d5bb21dc9e01e6ad63316fba71a688e89de86e04d62eec06421c8a9ed73e15",
-  "source/cameras.json": "e90788914897b93b566f8ba7eab738ebd0ab69ab9ad1a4f72cf196e1d371bf7c",
+  "source/cameras.json": "c71fa4909795a39289bd412017f3b84b22318966e9c7948a3cb5d5fef65d1a6b",
   "source/caption.txt": "2cfd291dd5ae0272a0e90607a3d059445111fa620a40d5323ac07d26a241660c",
-  "source/meta.json": "04763a07dddb539999143744249adf5023c7cf267de07ff7a76b03105a68852b",
-  "source/points_target.npz": "eeeb1eae8471d1ad277941c44a60b54ffab932d6e0c6bf1033776bb352f89b53",
-  "source/visibility.json": "a7269948a0bc263851bbbeecd333f32409d8db0de92a79ed55605fadd66a7892",
+  "source/meta.json": "ebaf401b395f269b4abcc86e79a352aea82e592ee3444c2a1fd55bfa042ca4b5",
+  "source/points_target.npz": "77ca0dece1e456f3a0b20adde532d5bfed1d9adaa4c4525e423075fa231df9bc",
+  "source/visibility.json": "289214c78d3b639cc020aa8db6fdf4fec35f8a825dcde7100ebac64d219cc07e",
   "tracks3d.npz": "98ae99493d4190dbe738b42f9d3e3553108b51cd9bccb87a8b70fff59220eb6f",
   "videos/view0.mp4": "2181ab7129515e3ae5519e6438fca840685689b5e21a5c1c15ac1762b4260718",
   "videos/view1.mp4": "e17c3f8810de65e94d3b9ce24e872b2854785b66a334fe5e9e3d271f08787b65",
diff --git a/Simul-Robot/c001_episode_006390/source/cameras.json b/Simul-Robot/c001_episode_006390/source/cameras.json
index 75741d85a6b239734f3579af643abbcacdf7dea7..580c16d8af749b85ec33da87e62f827c662199fe 100644
--- a/Simul-Robot/c001_episode_006390/source/cameras.json
+++ b/Simul-Robot/c001_episode_006390/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006390/source/meta.json b/Simul-Robot/c001_episode_006390/source/meta.json
index a150494352677765709739de015bbcdb6f6e7389..0e7570303feb459e51deec660a64145a5a3376ab 100644
--- a/Simul-Robot/c001_episode_006390/source/meta.json
+++ b/Simul-Robot/c001_episode_006390/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [11, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [11, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006390/source/visibility.json b/Simul-Robot/c001_episode_006390/source/visibility.json
index c9cc0cc39056b4200a95006434833d8f483294ce..e35b05a44f678f337881c5a8a67562ba05312177 100644
--- a/Simul-Robot/c001_episode_006390/source/visibility.json
+++ b/Simul-Robot/c001_episode_006390/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_seasoning_shaker": [
@@ -14914,8 +14914,8 @@
       "target": "blue_seasoning_shaker"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [11, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [11, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -26067,5 +26067,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006450/sample.json b/Simul-Robot/c001_episode_006450/sample.json
index ce40efc714d35d01d48851c03cdbe11fc1f7f594..33c5e1fa1128484e881dbdadd4a21cdbf4392023 100644
--- a/Simul-Robot/c001_episode_006450/sample.json
+++ b/Simul-Robot/c001_episode_006450/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "444b7fae2064c97f2157e8ed307aff2223d7933a8c9e0047a992aec91cc193b7",
   "labels/objects.json": "93477b0385d3ed83d776eaf1aa582a306ac5c47a12a9b7c885e1141d562db310",
   "labels/visibility_states.json": "5c28d63297393526e89552d5f0a8f9b1821af434bbf6548afd262743f6a889de",
-  "source/cameras.json": "65f6b9f6fe886ff42748a16324affed90eb96261e8f5e1bdb0932e0e1d09e365",
+  "source/cameras.json": "852e9582465c9c5163657bca54c270ebfa120c074b74a6068673a0f835d88917",
   "source/caption.txt": "d149b27f4a4d4208fa5ad754c1ab0ebcd453b1882409d7459ac38e9afc493d4b",
-  "source/meta.json": "ab48076a4731e29d7dae2c268af686c31a4e6136c376800dd0a2bfa607810bd2",
-  "source/points_target.npz": "3abdf9677db317496fd57718fd73904972a19c65bba522ffd44420204eb40cff",
-  "source/visibility.json": "a837da64c24591b4911151284508c7cc724531ecbe3f0614c01dc651ea69075a",
+  "source/meta.json": "5ecdc46aa061db10b643012db8ae159be2c3e588fba62c5176d1d6ea331bfd62",
+  "source/points_target.npz": "f4631ec2e8e0baf190e08a33cc8a358397b321f8eec68403e249257cbca769fc",
+  "source/visibility.json": "b8d30237beae2bfad51bc4e609c5990270fc4d7be9006eb0fc169630d70ea5ca",
   "tracks3d.npz": "9dfd79271c2dafdbcea828781b8ad15b525d9a8531ba369e924a3c3ea9e9e546",
   "videos/view0.mp4": "ccc445116ec4358330d042866c3063089c39e7ed2ced57298508380cb387339a",
   "videos/view1.mp4": "37976a2fc92f634e7349b1519426ea46d8f69dea8ab254ad5f80fcfe9b818614",
diff --git a/Simul-Robot/c001_episode_006450/source/cameras.json b/Simul-Robot/c001_episode_006450/source/cameras.json
index 4867c27264561a8ad2973990c0b06dfc109d27a3..b4a2ba01697f919fca2872f20106f02e291883af 100644
--- a/Simul-Robot/c001_episode_006450/source/cameras.json
+++ b/Simul-Robot/c001_episode_006450/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006450/source/meta.json b/Simul-Robot/c001_episode_006450/source/meta.json
index 0ffc736b1c2f6b9552599e110dc66305e77f2a18..82a1277101d05556d220b490b65f479d3f8d9bbc 100644
--- a/Simul-Robot/c001_episode_006450/source/meta.json
+++ b/Simul-Robot/c001_episode_006450/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [3, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [2.41, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 4f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [3, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [2.41, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 4 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006450/source/visibility.json b/Simul-Robot/c001_episode_006450/source/visibility.json
index e8917b2fd3393b0764eb9a0cb8104804167a0ed8..930a21512bdcd11cf36159011273950e5dadd81d 100644
--- a/Simul-Robot/c001_episode_006450/source/visibility.json
+++ b/Simul-Robot/c001_episode_006450/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "glazed_squirrel_figurine": [
@@ -13354,9 +13354,9 @@
       "target": "glazed_squirrel_figurine"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [3, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [2.41, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 4f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [3, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [2.41, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 4 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -23338,5 +23338,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006510/sample.json b/Simul-Robot/c001_episode_006510/sample.json
index fafe1d91c87b03d9c0ee4774c63cded89f79d561..177b7d52906d8f7c780c929c6def2332dd6b60c3 100644
--- a/Simul-Robot/c001_episode_006510/sample.json
+++ b/Simul-Robot/c001_episode_006510/sample.json
@@ -220,11 +220,11 @@
   "depth/depth.zip": "6318495672c95d2a15e252221a4143863c901bd0d6e88411b9a7973a0148f573",
   "labels/objects.json": "c2e1adb20848f266daa52dbb22e7f0df4f234aa66582f651902e0ecdba81afbc",
   "labels/visibility_states.json": "5a381ddeec2d7f554c2e4a3643cc7f9991ad13f1426a25acb373f01d22983af4",
-  "source/cameras.json": "f6fddac20cbe47168ab88a9a91eb239bc2e27d0cc1aa2324d01f16e1a010daac",
+  "source/cameras.json": "63900a07168e97741f0511b0efa87388c23b2be5371c642652bd9cd73f525dc3",
   "source/caption.txt": "4433c2cc6741fb8ade6a69c02137c42da7068b0997214df1a5e2da228dff4f9a",
-  "source/meta.json": "571ac7a9792d46d5bee3dad4fe5ede6700ec072170b1a06405c8f32d867a3825",
-  "source/points_target.npz": "bb23498a5a03a556453a56f68a2256811393462cb3cff9aa8e11e267b2574488",
-  "source/visibility.json": "3998de6ecd7c947677e356dfcd5e748f847aa39980ec00adc65c8e745be12101",
+  "source/meta.json": "9028e04ac7fb213406efb6b4d247def9a8678b3ba88ca46bc87fa6e8efcd2cd5",
+  "source/points_target.npz": "c236f3bd67e07e7f618abc03a65d9fac55c5add98a82bb1e757b62342988c205",
+  "source/visibility.json": "1d048f576f21584f7b87d09c5fb23c2bed75c5644762c49185788aa4822d1b06",
   "tracks3d.npz": "1edff1cc95e6105efa5319bd019db8a7b894af4869c22a0d4d7e24068ba40121",
   "videos/view0.mp4": "b4228e80730337698237d528c99e4cbc8b4d8927c1f47eefda7949b166fc37ca",
   "videos/view1.mp4": "7d915aad3d5bc35ec8fd3a634c9ba11d2b6084fa3bcd12a74be8c98f1c65fd25",
diff --git a/Simul-Robot/c001_episode_006510/source/cameras.json b/Simul-Robot/c001_episode_006510/source/cameras.json
index f1c1e2dfc80d4299e01c0adb0e627757f7d02f19..7d940ad25087be50548556cc27c36ac964355d21 100644
--- a/Simul-Robot/c001_episode_006510/source/cameras.json
+++ b/Simul-Robot/c001_episode_006510/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006510/source/meta.json b/Simul-Robot/c001_episode_006510/source/meta.json
index a984ac4b8f5eccc48323bb330b855a70dfec8a40..fb269a66b3d3466ed420fb5bb4c86499a4802ba5 100644
--- a/Simul-Robot/c001_episode_006510/source/meta.json
+++ b/Simul-Robot/c001_episode_006510/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [1, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 1f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [1, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 1 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006510/source/visibility.json b/Simul-Robot/c001_episode_006510/source/visibility.json
index 4a1cffa2acd0948f91a662f3293d993f5315fe6c..4c75920ca4cdb51711d49eea1aad0a4fae1b3f90 100644
--- a/Simul-Robot/c001_episode_006510/source/visibility.json
+++ b/Simul-Robot/c001_episode_006510/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "magenta_plastic_crate": [
@@ -16753,9 +16753,9 @@
       "target": "yellow_dog_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [1, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 1f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [1, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 1 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -29287,5 +29287,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006610/sample.json b/Simul-Robot/c001_episode_006610/sample.json
index a5b2056e263caacd0fc430d2708c12b66eff52f3..2110962765d8452a13c7becfad793b74ed9a05a9 100644
--- a/Simul-Robot/c001_episode_006610/sample.json
+++ b/Simul-Robot/c001_episode_006610/sample.json
@@ -220,11 +220,11 @@
   "depth/depth.zip": "795c771eb4dc7bab54e492095e1bb7b8eb6fe164711ec0d3dec095ef054aae63",
   "labels/objects.json": "be47fea3402d121846993fd7da8aef1f2e65c06a7875bc9f98f3012b0b12c888",
   "labels/visibility_states.json": "b2aebd6e588960c50ba4580a38f6f4a72e48bbdf73a300968b67ef12198ad634",
-  "source/cameras.json": "35bac38ca71d0fb88dfe129f6e6582db342ca0d44d96e497d1981ee661babc7b",
+  "source/cameras.json": "e595ebf770325fde1f7e6fc5f3cbf8f2003614190181fa0ae57bdbc9ff166c9c",
   "source/caption.txt": "b1bfc8bcc2ce3b40d997406612080196cd29615432ed7aabc10c1a94dc58ef6d",
-  "source/meta.json": "d6f7a98e980600c1a23ec9670499c2019197f5a2995d24c28860dd92790d4198",
-  "source/points_target.npz": "4fc383384051cb4e95b1013e93d979dac769f881333c7d2ce2af8f499ee6526d",
-  "source/visibility.json": "8314a43ed90bd0854fa446f41b250c7d5813643c8f36ee18a58b6c05c80b2a6e",
+  "source/meta.json": "6e1b059777e61a7f767fa99e72ac975a1f7322b8e7b84916cda3944de55810a7",
+  "source/points_target.npz": "8caa7e2a7866efea098f70bac07172428f71056d4c88e13af34ebba09a95ecff",
+  "source/visibility.json": "4c779c8d58bdc2bdf40d2285902b835e850aceefece86078982abbd905f623bc",
   "tracks3d.npz": "a7c54f4b9f7edfba309b26036e655868e5a637a924f8c4d1a93be868131c0cea",
   "videos/view0.mp4": "b57dea321b52c6c5869acae9de31c851898c14d05ae48c53df9147e9ebf504ed",
   "videos/view1.mp4": "2dd8241bfde2e7ebb73db0eb4ea88489cfb7910af2d356d184935686b304ad5b",
diff --git a/Simul-Robot/c001_episode_006610/source/cameras.json b/Simul-Robot/c001_episode_006610/source/cameras.json
index 9daa0259c360e0a57413369323860701f6a3c8d9..b3834c823382d0c8c471557e3b71a47146116ee7 100644
--- a/Simul-Robot/c001_episode_006610/source/cameras.json
+++ b/Simul-Robot/c001_episode_006610/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006610/source/meta.json b/Simul-Robot/c001_episode_006610/source/meta.json
index 7a39948655713ecb74e958407c06639413875318..afe5aea4046dfff5a65efe6d2cb3659a671d3343 100644
--- a/Simul-Robot/c001_episode_006610/source/meta.json
+++ b/Simul-Robot/c001_episode_006610/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 1, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 1f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 1, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 1 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006610/source/visibility.json b/Simul-Robot/c001_episode_006610/source/visibility.json
index 931c3267a30c16f005d97a49323736c3ebe26e6c..e7e1e34114a4820c5eddf99957738bd161546c96 100644
--- a/Simul-Robot/c001_episode_006610/source/visibility.json
+++ b/Simul-Robot/c001_episode_006610/source/visibility.json
@@ -48,7 +48,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "peas_food_can": [
@@ -26474,9 +26474,9 @@
       "target": "tomato_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 1, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 1f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 1, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 1 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -47612,5 +47612,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006680/sample.json b/Simul-Robot/c001_episode_006680/sample.json
index 2c92a12ab7ff14c4f06c21821f51de0b6e239db4..6c4a7d73f9ec84b49b0d660e448f9ca6263d959c 100644
--- a/Simul-Robot/c001_episode_006680/sample.json
+++ b/Simul-Robot/c001_episode_006680/sample.json
@@ -218,11 +218,11 @@
   "depth/depth.zip": "c8f8a2951e2156bb673cd0610a332d80f5df26c6b1fd4e1eaf576c3b7da19c86",
   "labels/objects.json": "c7ef94a43aab676570c61f17a01aadb5ccc020145f65eea033f37232e3b3fae8",
   "labels/visibility_states.json": "d56fb1c7d2dac1dac5d4f2a33c8d0b36bd646449d1f0fda56f8e3b8c93ee0946",
-  "source/cameras.json": "04a75f8d0f84cec5ecefa46224f4f093d7d67334c9c9e67c076dd2a87a7a0112",
+  "source/cameras.json": "5075002618f5252d6240287ab909347f30da124b3082e864689512d87259dede",
   "source/caption.txt": "b9683550aedc00145d1ecefe1ec7ecccf107673f366abaaea5e2f9ae3b2d427b",
-  "source/meta.json": "47dcaacde5b9072531a5fe8e7ff878b9a4aaccefe59cd0285ffdc82ca2c9477e",
-  "source/points_target.npz": "5c0fb99e9fb7933cabe50a36bfd40000549aa6cd0fb77654a5b6014e49990cb5",
-  "source/visibility.json": "1286f781ac078965d5b496324403a59ad60fa9c59fd34ef9a7c18a5cb7ef259f",
+  "source/meta.json": "610fc692e718905bad7166e3bebe52da8be791a3f87146fbd8a78fad5537d347",
+  "source/points_target.npz": "13b33d312021d4728c11f74a0b2b0cb2f05713706d836ad7be060387fb6f31fd",
+  "source/visibility.json": "c5702f2e2bc23f378c8948b7433eb7fc9824506c5bfcdcefaf35b1bae8153173",
   "tracks3d.npz": "b823d7220ca084aad6812745bedfcbc6bbcbce31fe9f522af26268b84a77942f",
   "videos/view0.mp4": "15c5e0d178bf5b4856c813a273b74cff61c230a3f705972d3b043163f72053c9",
   "videos/view1.mp4": "b9cf3e9d1731a91b0fb42d5d665639046b39c31be3e9848433f441e04760c0d4",
diff --git a/Simul-Robot/c001_episode_006680/source/cameras.json b/Simul-Robot/c001_episode_006680/source/cameras.json
index 002d8f43747ee852d9468a5aeb41279282815844..30b89b2cc8a12356c2b4d99ac56ba0cb7903c061 100644
--- a/Simul-Robot/c001_episode_006680/source/cameras.json
+++ b/Simul-Robot/c001_episode_006680/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006680/source/meta.json b/Simul-Robot/c001_episode_006680/source/meta.json
index e6380ef821188a6778b9794e24829e8bc98445d2..ac1fbf9887ace415367e6be3901108ab9f6efd03 100644
--- a/Simul-Robot/c001_episode_006680/source/meta.json
+++ b/Simul-Robot/c001_episode_006680/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006680/source/visibility.json b/Simul-Robot/c001_episode_006680/source/visibility.json
index 4190c3aa7f2d02f607bed2b4a24cbb2dc6c9f64d..9b680bb5eaa0cd3c943265761ff9bb7468f956a5 100644
--- a/Simul-Robot/c001_episode_006680/source/visibility.json
+++ b/Simul-Robot/c001_episode_006680/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "painted_tin_can": [
@@ -14110,9 +14110,9 @@
       "target": "parmesan_cheese_canister"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -24664,5 +24664,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006690/sample.json b/Simul-Robot/c001_episode_006690/sample.json
index 4a1289caca71b026d844243ef842df7c0f4be4da..b5a37596747b40801f7b48463dc5e02e0c635f7f 100644
--- a/Simul-Robot/c001_episode_006690/sample.json
+++ b/Simul-Robot/c001_episode_006690/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "f9e2bf861857198bc15ea64fb14f289123fb39fb5a7d258660300419f4389225",
   "labels/objects.json": "50a96d9c8787f85581860b947841e7ba861b8a16e4d4878b6dd24fa9e623502c",
   "labels/visibility_states.json": "fc698374aa0a8f110ce724dd7ab4fd5c8195969778997916a30f2bd346f55e8e",
-  "source/cameras.json": "b0086a2b83d0398a1fdd6a4fad29e7c7dedee969eb6a79051847d0c46b2aed46",
+  "source/cameras.json": "76e3f9af9868798269439bf8ca42de50b06f767cc245ee9b4e3f51f032b80a19",
   "source/caption.txt": "92a936720ff9d1d747af71165e2886f5aa18a0d26023b822a7ad5af26d9409d5",
-  "source/meta.json": "ca43f6891c6cf6d764f2d309160c2203b1a51597c4a506053c6197be280f26ea",
-  "source/points_target.npz": "c5e37530f6744ff8f22d269c2f1660fd19748a30b45a54d46e90fcba69acdb95",
-  "source/visibility.json": "e22e8f243d868442ad55c7c0c7d214b326554be19d274b2b955973f8968a9af9",
+  "source/meta.json": "dbe5ee00d43f9a8c7873a4e02ecd886c4377c6393176cdd43fb25fca84b09b35",
+  "source/points_target.npz": "e7e3cf40c63c27cf347f0eef55893bc54e874cdd6411c1089ef7e61bd1aad37f",
+  "source/visibility.json": "34c2eb884441d44de79b94aa5d81e47d85756ef014fe4237345ac7cdcec023a2",
   "tracks3d.npz": "14988d4889855fb951658a3e9f918a445bbd457c325d42df9f77ea8a356ce0f4",
   "videos/view0.mp4": "6596cd1d24860b03efaa323312dd20567772c3ca6a639a0d9a88812e725e9512",
   "videos/view1.mp4": "6c7002d0b338df2e11e098829f96e80f7633e03eafddace5ec11c123fd0e06c3",
diff --git a/Simul-Robot/c001_episode_006690/source/cameras.json b/Simul-Robot/c001_episode_006690/source/cameras.json
index b617a71457fbffb512201f9b0004e734fc61099f..c4cb0df18fb6dbe9f4ab174f78dc3e2293faec24 100644
--- a/Simul-Robot/c001_episode_006690/source/cameras.json
+++ b/Simul-Robot/c001_episode_006690/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006690/source/meta.json b/Simul-Robot/c001_episode_006690/source/meta.json
index 186f8f8fe9e2517185ef74b69914a8110de7d648..e6f30d76136718261fd42e3c07045ef4df85f957 100644
--- a/Simul-Robot/c001_episode_006690/source/meta.json
+++ b/Simul-Robot/c001_episode_006690/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006690/source/visibility.json b/Simul-Robot/c001_episode_006690/source/visibility.json
index 1b268755a55c048fe5f2f89238410a8b83077be1..2a777184305274d5bf6841409076f10147e643c6 100644
--- a/Simul-Robot/c001_episode_006690/source/visibility.json
+++ b/Simul-Robot/c001_episode_006690/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_seasoning_shaker": [
@@ -4168,9 +4168,9 @@
       "target": "painted_tin_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -7466,5 +7466,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006840/sample.json b/Simul-Robot/c001_episode_006840/sample.json
index 5bfe5dbc3a3865700839636495ddb3d95ad14db3..2d280faba966ab2bb5e0eb78af5455568a5bb8ba 100644
--- a/Simul-Robot/c001_episode_006840/sample.json
+++ b/Simul-Robot/c001_episode_006840/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "b000cd3166707eaa45433b72112cd69a03b1011fa750244fc079120198a18565",
   "labels/objects.json": "5dde56bd9696229032052aecdc0685071dde0dc8ee759097358fc318250dce05",
   "labels/visibility_states.json": "6f3d3a3bf951b5d7447ee5fce7d076d7cfa8fce49f5a19d8ee94b4332fe7a475",
-  "source/cameras.json": "d1efae5221f4daa9c0f878675e036ae55d0fc288544077f670229296f1136e24",
+  "source/cameras.json": "03e9dd54d4e271fbe1c97dc8eaab178655483597410168fd183495eae5b7b628",
   "source/caption.txt": "6bda0387afd9e4aa926891ef56668282a2a5e650e04117af28100dd92d36fe57",
-  "source/meta.json": "4199d063568790bfdd5d5414f7a1c1eb3309b9491038606eed785dc6e1733357",
-  "source/points_target.npz": "be2984f3fefddf81b75c9616d258bd4fae13e35c0263d22c27678e62d0cb72a0",
-  "source/visibility.json": "3be0dcfb7dfb907f2f09c4abd808660b0b05f2cfd27255acc4a8c647457ee2f0",
+  "source/meta.json": "3e1ad543bf89098460f4546b7c5a6f6fcfc90f09233656ef6d61916f54485013",
+  "source/points_target.npz": "b8d18fb56f6b2b84368a782bb7ffc48050c806bbf2b70afd74ca08626e1e6b09",
+  "source/visibility.json": "31249eb2d84a4371b53b25c7ddddf3df99718602e9bdc741454c012deff016e2",
   "tracks3d.npz": "74a8781b215249ce3b957dbc2fad26e4d75cce43a5f2e196d0268d3dc888d005",
   "videos/view0.mp4": "33bfd25610fca16f447b1b76ff1a95eaca0acbeab77cf78f864689f71ea88285",
   "videos/view1.mp4": "7398e572679c9389897c9b498f3797eb5f5dcf9be3b29f9210e9ee96c3c6fa1c",
diff --git a/Simul-Robot/c001_episode_006840/source/cameras.json b/Simul-Robot/c001_episode_006840/source/cameras.json
index cd475cabd3f52a680ce7c6aff9db0ff67b0a67be..3c5827025fc11e1139323f9dd7409079f978ed8f 100644
--- a/Simul-Robot/c001_episode_006840/source/cameras.json
+++ b/Simul-Robot/c001_episode_006840/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006840/source/meta.json b/Simul-Robot/c001_episode_006840/source/meta.json
index 7068c5954450686ae08051f05fd9f6985c3e428d..91833c2ed87222ca88374ff83f0710bc37599caf 100644
--- a/Simul-Robot/c001_episode_006840/source/meta.json
+++ b/Simul-Robot/c001_episode_006840/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 20, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, 0.04, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 20, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, 0.04, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006840/source/visibility.json b/Simul-Robot/c001_episode_006840/source/visibility.json
index 6788490ca5ac0258e6a21fc85f61b72440e598c0..b0ee4184a1770e648aa15078d7820035e61e7436 100644
--- a/Simul-Robot/c001_episode_006840/source/visibility.json
+++ b/Simul-Robot/c001_episode_006840/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_energy_can": [
@@ -17314,8 +17314,8 @@
       "target": "yellow_energy_drink"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 20, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, 0.04, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 20, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, 0.04, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -30267,5 +30267,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006850/sample.json b/Simul-Robot/c001_episode_006850/sample.json
index 93578880a95fe4a944c75dd07746d8883813ec51..c131a6f71aa361fd9e862b80c6482a6e00fe0d30 100644
--- a/Simul-Robot/c001_episode_006850/sample.json
+++ b/Simul-Robot/c001_episode_006850/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "b7ef5af643ef6cd16946f7b9a8038997283d16c4096a38a4201184e311b53a09",
   "labels/objects.json": "fa2cb7b3ca801fd7676397c825b394e802d5ee447255ab67ecb5eaddcff5cea5",
   "labels/visibility_states.json": "b9fed7d572ddbae92f45b7f352cbb997ba455b3d9fe07833aafa0f9450c376e6",
-  "source/cameras.json": "1428853b859b2cc77e3b81ce0ae46ce2308e77d0aefc47faafd4828b11a863b8",
+  "source/cameras.json": "4d358c2918a46a192f11dde9e79ce54cd394382e409d34e7e802cbf21fb2d778",
   "source/caption.txt": "d7b41b106e482e4fa470d30957d7e5d63e18fa562ce60406f58b825ff1bb68b7",
-  "source/meta.json": "303db32ccfd72b11ffd752479941609f5d04994832957a64ee47e551bdf546ab",
-  "source/points_target.npz": "fbf196eb575bd549290380077556b7b1ba63ed90090c474af2b3247e58422227",
-  "source/visibility.json": "722bf7e2438eebf88c4a3663967897859764e2231244946c6128af28933d449e",
+  "source/meta.json": "9320a8b6364620dde1f16f1ff0d18d4fd40a92ab9404ff827f4255f890befbf9",
+  "source/points_target.npz": "f7ec65aa1a6ba32cd74c98352c724bc027ec03392e8c2990698e098824f6717c",
+  "source/visibility.json": "9706f45545ef478f459ed65702760ad61fd2b6c70384f106a3c0a7e304078fc6",
   "tracks3d.npz": "b4ff149b2537764fff2834a78a51e7acd997f5084cab6454bde3f3fbc22f0033",
   "videos/view0.mp4": "9a1e0c66b20f98f21829ac589b11bc32cc224707559f5e26d6ecf1e5e878ab1a",
   "videos/view1.mp4": "69accdd78d023ba4d7927056da9382a4d0786c79b4f1a9c31a5ab397ee0dac79",
diff --git a/Simul-Robot/c001_episode_006850/source/cameras.json b/Simul-Robot/c001_episode_006850/source/cameras.json
index 95ce449bb816ce32b348d4af5559919c49e5a011..6e0cd6f055972cb4a16e45b642edaf7140ca6b9a 100644
--- a/Simul-Robot/c001_episode_006850/source/cameras.json
+++ b/Simul-Robot/c001_episode_006850/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006850/source/meta.json b/Simul-Robot/c001_episode_006850/source/meta.json
index 6e326611f337c8d9df407576be4aeca7316cd158..61ad7002bec54e8db723409ac14770b0f3dc1326 100644
--- a/Simul-Robot/c001_episode_006850/source/meta.json
+++ b/Simul-Robot/c001_episode_006850/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [3, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [1.08, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 3f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [3, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [1.08, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 3 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006850/source/visibility.json b/Simul-Robot/c001_episode_006850/source/visibility.json
index 2f3ddca9b9414df3600c4eadda71634829f87ce8..4a418dbc7df0473572911f06a827b60918f4bb93 100644
--- a/Simul-Robot/c001_episode_006850/source/visibility.json
+++ b/Simul-Robot/c001_episode_006850/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "small_red_cube": [
@@ -5121,9 +5121,9 @@
       "target": "small_white_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [3, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [1.08, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 3f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [3, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [1.08, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 3 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9179,5 +9179,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006860/sample.json b/Simul-Robot/c001_episode_006860/sample.json
index 1718c754f5efd63fa64e7b0ba85222a8d13c35e4..31b1cb51fe111fd7b1141ad17d6e4c2aa65139c4 100644
--- a/Simul-Robot/c001_episode_006860/sample.json
+++ b/Simul-Robot/c001_episode_006860/sample.json
@@ -218,11 +218,11 @@
   "depth/depth.zip": "e89a60f7ede1431cb41d511fe2f64b103618160ad14fb5274b9775ba5833ae1e",
   "labels/objects.json": "25393d212bb7687303eb8a8cc942ead3352eb1e17422b83a2c1566343cbca344",
   "labels/visibility_states.json": "aa3e3d740fc9fd1f8dbae75e6b17c77ccb524de4ff04b1b0aae233ffbecdb305",
-  "source/cameras.json": "ea98ea217219ae987d0bcaefc18746ec7c57fca069eb9c4e74180267c67c92f4",
+  "source/cameras.json": "f274a113e0e954018729c5162dcf60115e2f5a4585cf6fd74b0f7caba1dcb33f",
   "source/caption.txt": "7c01d5b9e13f089a566376a626d740e92b86873815e8144e6a4ff9c7a43137dc",
-  "source/meta.json": "2c282be1f1bad418a2b66eb6490fe1e30c4f4ae2449173498a9b07f0ae25e853",
-  "source/points_target.npz": "d674b37d519ff83dc0fecb189bff12ec396896af90dd71ed3067b30d5acef74d",
-  "source/visibility.json": "8e14bd82f8350c830d8a8e13177078b7a061554fe3325ecc8dcd39c2de0f0fa6",
+  "source/meta.json": "0bf8e132149f77f1dbadc3d12f0bfcf36ed8d1b0030818a113c7bc917a56479c",
+  "source/points_target.npz": "841e4b41484d940a8c23bce3865183d93f3f5e0887f71cca8f8151454aa3ce7b",
+  "source/visibility.json": "fd0eb0191ee3d15c8e1bdb9f9dce8316b5f9f6c77e92c20ddabd67ac7ae09cb5",
   "tracks3d.npz": "5b7cdd7af7263a45ab6397214e5b5815a2de87935d0d2cdc950de672f00d6490",
   "videos/view0.mp4": "709943adfb3288b1e9b52d450a2245b743fc1bd1ebc3db39feb0c0db69f47578",
   "videos/view1.mp4": "1585ec8d470e6530e9cd462a37d1a9064b17d79b003d3a2a6520363bf4bd96c5",
diff --git a/Simul-Robot/c001_episode_006860/source/cameras.json b/Simul-Robot/c001_episode_006860/source/cameras.json
index c6bb8b0cacfe01b8522c8b45b6b399c5ee630af7..b76db87e911dd9002e53b16881c9a10e554183ea 100644
--- a/Simul-Robot/c001_episode_006860/source/cameras.json
+++ b/Simul-Robot/c001_episode_006860/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006860/source/meta.json b/Simul-Robot/c001_episode_006860/source/meta.json
index ca425251cb26c2d6e3f14a22fa97335c29912c56..7fb092a90063f22484347156951f998b1d9effeb 100644
--- a/Simul-Robot/c001_episode_006860/source/meta.json
+++ b/Simul-Robot/c001_episode_006860/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [14, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 18f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [14, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 18 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006860/source/visibility.json b/Simul-Robot/c001_episode_006860/source/visibility.json
index 71c183bec6626ea690cdc773e9ddb7db62be92bd..d9b98a0b67c2de24b171867b7e470b897b0043b5 100644
--- a/Simul-Robot/c001_episode_006860/source/visibility.json
+++ b/Simul-Robot/c001_episode_006860/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "black_energy_can": [
@@ -17393,9 +17393,9 @@
       "target": "black_energy_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [14, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 18f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [14, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 18 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -30407,5 +30407,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006920/sample.json b/Simul-Robot/c001_episode_006920/sample.json
index 0112e44264f0f7a590ad8dce84c5d0321cce95ba..826db11c8807445b56ad9adbf3d89b054a567f2d 100644
--- a/Simul-Robot/c001_episode_006920/sample.json
+++ b/Simul-Robot/c001_episode_006920/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "9c1902b98a27bc3eff67b4ed04342d02ac8d3f20b5d76e99ad16c288fd6eca68",
   "labels/objects.json": "0b7b7fdee9bacdb94bbddbec92a1aa48cfee3cc1b82f2fe7232aa2a58d7757b2",
   "labels/visibility_states.json": "acf9aaa6e4fc545833222e92276e297c8012e4e679dba35b349e87bc6192cbcd",
-  "source/cameras.json": "a724550a7ac1e1963837b425c4114d38aa06f06e1770e1f5e8279ace9d3d6fc4",
+  "source/cameras.json": "a587690f75f3443421dc323b2a837ae18b4323c953a5ecd9daa87ad54783b500",
   "source/caption.txt": "b6b7cd40a75a2d543bf3f8e69198a4db1c432773fb0ae566d049a5dcafd103ae",
-  "source/meta.json": "9f3762b34b061076c5a842408db9f121e65d0551031e6d98ed2b10868eb9ae3e",
-  "source/points_target.npz": "9db6289d4a06274216fb5bffb9f2e2443cbcf31821b2b0c99311079b132a5810",
-  "source/visibility.json": "fd61400decf0b170214b4e7dbc8551f3b68ff6ca92737ee20d25ba3d0123371a",
+  "source/meta.json": "9cafff3c8ab363419ca2331992ff2f584c9b95539a0b786525c6c8ddb29badf8",
+  "source/points_target.npz": "25ddeefb3e692e8dd67efb4c4ab607c627dca1b20a7b3eee1c216b0d997e991c",
+  "source/visibility.json": "caa5bd986c98b4d33f57c32ab09c256780ffe6e9e5ef945ad295f5059f259162",
   "tracks3d.npz": "da01de45b1259c983c30b3f0d170ee377836522e26b78ea52dfdb5b763cb2d3c",
   "videos/view0.mp4": "e8ddd505a0da36408f51502d0d43e9abfe911bb4b078ed4e4c1264e1436a8822",
   "videos/view1.mp4": "d5e538d4a8187b3f124f2015ee2642ad1574283bfc7d116731319901d4336f9f",
diff --git a/Simul-Robot/c001_episode_006920/source/cameras.json b/Simul-Robot/c001_episode_006920/source/cameras.json
index 82bf770409b06d0f87b7ebf6465287cb612edff3..7c2d41359054fb024162003c5b46fafd5f1ac044 100644
--- a/Simul-Robot/c001_episode_006920/source/cameras.json
+++ b/Simul-Robot/c001_episode_006920/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006920/source/meta.json b/Simul-Robot/c001_episode_006920/source/meta.json
index f41e10679f24cc1d2b0b131d0e6ecf6237486670..c408daf9ac397ec36f7174e2c7e9cb18ec158b6b 100644
--- a/Simul-Robot/c001_episode_006920/source/meta.json
+++ b/Simul-Robot/c001_episode_006920/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [12, 2, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.0, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [12, 2, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [0.0, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006920/source/visibility.json b/Simul-Robot/c001_episode_006920/source/visibility.json
index 4f66d1e25e6770410b686e64ffc77e3f80dcb8a1..a8bd89cdf3acf8077021116631d5b119e103c972 100644
--- a/Simul-Robot/c001_episode_006920/source/visibility.json
+++ b/Simul-Robot/c001_episode_006920/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "art_deco_bust": [
@@ -14397,8 +14397,8 @@
       "target": "brown_ceramic_bear"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [12, 2, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.0, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [12, 2, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [0.0, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -25160,5 +25160,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_006950/sample.json b/Simul-Robot/c001_episode_006950/sample.json
index d52be6490f10dd96b2181c7517f7bd610e29faec..13eacac4aaf0242f4cd396b4a57c30caa14c7bad 100644
--- a/Simul-Robot/c001_episode_006950/sample.json
+++ b/Simul-Robot/c001_episode_006950/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "f8afeed7abeac71b1d7d723ee7f24d238cfd8816822db88a4d9be3d5302c6fcf",
   "labels/objects.json": "9473d3f89e6dc58661b4d322ba01f7e7aa70bf943b0b3aa010e7ede23a178bb5",
   "labels/visibility_states.json": "955af824f3449e3d02a7fedb0958f9d1e6a643d47b65461b0867c2d3edbfb8cb",
-  "source/cameras.json": "3d4a38f82df2979d58c87a41929558d36a75ba2c6eefabf6b8275358c3115dc7",
+  "source/cameras.json": "0fee0532c09da57d3dcc7f7211b0cc92d266e90d27936f9450f2237a237a5c6c",
   "source/caption.txt": "969549e5dc137a3f08afdda387b99d6734e25ec9c51a9606a291de9000fbb25e",
-  "source/meta.json": "7f54960ebae32c0a47a3c036e9b3929938dee1667c4ff267311e825fc4fee274",
-  "source/points_target.npz": "7cb7bd3dce3619d8f7a2ad571dcd1a3e17df3c42dd59f6c291d2988069474678",
-  "source/visibility.json": "27f3b3b8c6be0f7a94a7b1b96fa00ee64d04cf25f053ed92816916632c629a2c",
+  "source/meta.json": "9169057fca7fd493a7dbb0bc33931658475c06c8939ef00f8141b197554077bf",
+  "source/points_target.npz": "3ba729e2ed2291532ea12b909daaf1a5a62613c406ec03ed576e06a090f5d9b4",
+  "source/visibility.json": "d07526962a8c0822bd8dbafe756608668c4214f0187a147fcc0b8403b1f2f8ec",
   "tracks3d.npz": "6c2d3906c31c632e7dd4e52b40a9288b1d7576bc5f9c9d08c239a33ab98732fe",
   "videos/view0.mp4": "67ac99c13b127d9ee57108bceee2082d8b37dfe2c87bfc6f4c1e62f0ddcc1d23",
   "videos/view1.mp4": "db09d186f733211bf97b6ae2e5f9e07ea26e5a4690ec56d4e09752ed90eb8640",
diff --git a/Simul-Robot/c001_episode_006950/source/cameras.json b/Simul-Robot/c001_episode_006950/source/cameras.json
index 73bcd0cbc547d6668c4062d5d3500bc046754d0d..09b692f4ae48b0201909ca35e55ad03213bd3a3d 100644
--- a/Simul-Robot/c001_episode_006950/source/cameras.json
+++ b/Simul-Robot/c001_episode_006950/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_006950/source/meta.json b/Simul-Robot/c001_episode_006950/source/meta.json
index fea12241870a8453b9bf65b82b1dab8f934010ce..28c8f19235fcbea09aaee25f08889599191849ef 100644
--- a/Simul-Robot/c001_episode_006950/source/meta.json
+++ b/Simul-Robot/c001_episode_006950/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [19, 0, 1]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [19, 0, 1] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_006950/source/visibility.json b/Simul-Robot/c001_episode_006950/source/visibility.json
index 33b496e88f07701c4e094d6b2eb859ce1ab2d020..855798242d865d6b88b6bab740edfb4abfc0980f 100644
--- a/Simul-Robot/c001_episode_006950/source/visibility.json
+++ b/Simul-Robot/c001_episode_006950/source/visibility.json
@@ -48,7 +48,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_plastic_tote": [
@@ -26626,8 +26626,8 @@
       "target": "small_white_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [19, 0, 1]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [19, 0, 1] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -47883,5 +47883,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007080/sample.json b/Simul-Robot/c001_episode_007080/sample.json
index 481a3ee8f57ec440d445f911d499641b0eafcbb2..6963673e9fcbda8f80982b65b8626c3f109cb0c8 100644
--- a/Simul-Robot/c001_episode_007080/sample.json
+++ b/Simul-Robot/c001_episode_007080/sample.json
@@ -218,11 +218,11 @@
   "depth/depth.zip": "81ae4f267e6a7b77662d1f9150be2ec15834e46d1256d82333bda96c0a565190",
   "labels/objects.json": "2c5c0eaeb194c4c41fe03d8cfd8f0816f439dbd95692c71fc5fde0d45ceda771",
   "labels/visibility_states.json": "390442482194a7901613c5b98a2e962bc235b4389fd2905d38d6f3c358999c8d",
-  "source/cameras.json": "f3d931f816f5cfcec8998c9302bf7d64fb736266ef9b380e15029174ba8cda5d",
+  "source/cameras.json": "0abf28748cf4849f4c1db69b412609a15bd54d72b09e7c15272758b2373996e1",
   "source/caption.txt": "031ab5f9bfcdaf936294e69066762f2316d570f597794d203eca83af28d3f5b1",
-  "source/meta.json": "08c9d70e8794b05a034d2cdc4d2788d455b6694ae52d5c0a3fe7dfacd35c8da2",
-  "source/points_target.npz": "403332d5ec22e2285eaef4b5010130299d50c0fd4775cc0bea7d9e3a7365ce32",
-  "source/visibility.json": "e6fbd148508a012af5101899dbcd88d3632995e950c2d590389e03ed6a0477d9",
+  "source/meta.json": "82faafaa2dffe1b5580bec082589f072a9fa26c794225ac3c691e184d742119c",
+  "source/points_target.npz": "1433458302c8800b59175ccb43d660a8432665b9071f803c8e04624e113717f9",
+  "source/visibility.json": "0fb7697ea28d70b293d9dbe5bdad070c3a200d77f31a23ffc7e141aa6b34edb0",
   "tracks3d.npz": "02c2fa18583c70a81b1fb600a67126c00a6919ce1afd3faae8056d830aef8cd3",
   "videos/view0.mp4": "60bc84d062749ade311a382cd6fa7d700f5449715d1b6be11ea7dd759d380d5e",
   "videos/view1.mp4": "c0084029d8c380255f0cda7260164895008ad669658cb46e218798ad4f57e3fd",
diff --git a/Simul-Robot/c001_episode_007080/source/cameras.json b/Simul-Robot/c001_episode_007080/source/cameras.json
index 5d3ab637700a534e7f495484a9c1b278829d9015..736eebbb2cd7a7d6d588b56e7efc16c5921c2a09 100644
--- a/Simul-Robot/c001_episode_007080/source/cameras.json
+++ b/Simul-Robot/c001_episode_007080/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007080/source/meta.json b/Simul-Robot/c001_episode_007080/source/meta.json
index 1bd6850f5f548a7ee186059f2a7a57d5c2981834..573c2bce81a0292dfa25a9c13869907ec5f860f4 100644
--- a/Simul-Robot/c001_episode_007080/source/meta.json
+++ b/Simul-Robot/c001_episode_007080/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 2,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 27]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, 0.02, 1.93] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 27] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, 0.02, 1.93] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007080/source/visibility.json b/Simul-Robot/c001_episode_007080/source/visibility.json
index cf041772a058a157f2c2330d8d2598a269088551..c171b41e7146eef4cf821b6aaa386f00be84e5dd 100644
--- a/Simul-Robot/c001_episode_007080/source/visibility.json
+++ b/Simul-Robot/c001_episode_007080/source/visibility.json
@@ -42,7 +42,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "magenta_plastic_crate": [
@@ -14356,8 +14356,8 @@
       "target": "small_green_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 27]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, 0.02, 1.93] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 27] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, 0.02, 1.93] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -25089,5 +25089,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007090/sample.json b/Simul-Robot/c001_episode_007090/sample.json
index 8b4a03c9852e114fb6cfd4c07c40fd261018e276..adb1c08cf290f546e327d24e8ac0ab5fe651965d 100644
--- a/Simul-Robot/c001_episode_007090/sample.json
+++ b/Simul-Robot/c001_episode_007090/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "97da05487da56b1869ce11cdbe69ea37c6545b62395a320fbf26deadf09b9548",
   "labels/objects.json": "f30db0c0bbd0fc9021b8b8994b6a01644fb594e92ce6fa3b41468f2b744c378e",
   "labels/visibility_states.json": "a6cbb049b0c962bbc586b99a0937206f30254ec5d31d36f5b4871b1aaafffed0",
-  "source/cameras.json": "f2f519f59551f800620e5ac7804d718b999db0816238c5e618107aeba6238e48",
+  "source/cameras.json": "b22f59f29edbeaabb309c5264eb2de002fca963da27857f51f3a5fa31a43d9d8",
   "source/caption.txt": "d083ecb252ae3e5aa76e244a1ca1c41d5b21f91820a68d2823901ceb0a64dcd8",
-  "source/meta.json": "95fda64fe73f3d5cebfa7aa0172e1f46750239b2761a93758c62b15e708a9250",
-  "source/points_target.npz": "c123dab9f148ed76ce57f0cf3792f212bbbadb3d7ef0ec0467ee67e159f049be",
-  "source/visibility.json": "7e37c6627e335d72dc2a75fabd580fd5d1a9557019f8ff429e62f69101998ac8",
+  "source/meta.json": "6bdd53360cb1eeed9b5d99aed3d01094180a270d9c1a86ff1067327b3e13ae81",
+  "source/points_target.npz": "992e059f48dc72ca4690beac3a35fffaa2ba99b375dcb363060cf30819ed1936",
+  "source/visibility.json": "3853c211c52ee85f269751c0108c4274c0d48f188666dd99d53dc2a35f46a2f8",
   "tracks3d.npz": "d5a11ee878947942675fcaca19279b47b12a3faa4a58e67a6d195a52ae25adc9",
   "videos/view0.mp4": "9eb76fa3ce32e1910e5ba8dfba988875af2eaf771998870336a7d32e6031a830",
   "videos/view1.mp4": "039e4f4b84595174335d7d1057f1f6ea0a34935969daa23873119ad7077b6248",
diff --git a/Simul-Robot/c001_episode_007090/source/cameras.json b/Simul-Robot/c001_episode_007090/source/cameras.json
index efd3482963fab403a2c71d0f7865ec5a64a97b18..32f0679183d98c04c403285445d40e600e34c78c 100644
--- a/Simul-Robot/c001_episode_007090/source/cameras.json
+++ b/Simul-Robot/c001_episode_007090/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007090/source/meta.json b/Simul-Robot/c001_episode_007090/source/meta.json
index a804e20f6139d0c267d33f3389eb70bdc0900ea7..d97fa154c2134855e063bae3e257209a69d8e90e 100644
--- a/Simul-Robot/c001_episode_007090/source/meta.json
+++ b/Simul-Robot/c001_episode_007090/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [8, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [8, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007090/source/visibility.json b/Simul-Robot/c001_episode_007090/source/visibility.json
index 73010536a257dffce104ef4982ee0fc4b2a2301e..e7dfb7644d34ee47488d210204fe44e161cbd969 100644
--- a/Simul-Robot/c001_episode_007090/source/visibility.json
+++ b/Simul-Robot/c001_episode_007090/source/visibility.json
@@ -48,7 +48,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_plastic_tote": [
@@ -27823,8 +27823,8 @@
       "target": "small_white_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [8, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [8, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -50040,5 +50040,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007170/sample.json b/Simul-Robot/c001_episode_007170/sample.json
index cb9865ef06de72425fde87f90164108a1fad06af..e5bb73e83975f91bfb0fd2c3723af6658ad2e5f6 100644
--- a/Simul-Robot/c001_episode_007170/sample.json
+++ b/Simul-Robot/c001_episode_007170/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "93e95071f0882e15b6d55c4df3febf7bf315e7dcd96a3487b688e5b89201de88",
   "labels/objects.json": "0914ce192738c909db90fd47591abff36dd2829888d8151cd68bd7ac5a9edc23",
   "labels/visibility_states.json": "285eb3c76283254e7ca24dbc83670693495bfdf4e81bab538453fe0557e5268a",
-  "source/cameras.json": "33a25e2ed5a2d78fb1d6e38f52e43b9116b9fa5853d13d9fd6cec0c38195993f",
+  "source/cameras.json": "e58d67065557203d62f47f976545806cf0dd8adeac5a8bc0f556d2ac6a38b567",
   "source/caption.txt": "2644942ef2865c1b316dc54d3dea1c89dee61da5e493aa3c94ecdc1661c64223",
-  "source/meta.json": "67508cf5d71d5e02795030b7ad22ee931ca26faf83abebc3d98e6025082bd34a",
-  "source/points_target.npz": "d6c09b31def0c539b2e3e7dfcac8e0bab4b2efe580f47c7b2b8c9e28c933eab3",
-  "source/visibility.json": "8b0dbd17ec9e7eeb9c39e38cdc1a985df92770b3670e98857ef3d4926672c680",
+  "source/meta.json": "43860047dd6b0fa746b61f144e7457030298b7e37bc674e045c8c072e31b6e6c",
+  "source/points_target.npz": "d7134df99bae8167cc30dafa8e98781a3577e8aeebf09911026ce8eadb4c3350",
+  "source/visibility.json": "315ca04f388869ab808a511bab9ef630dcdbb1bd967da656f342e94ad1d3ae68",
   "tracks3d.npz": "9dbe79a431be6c790a2db9102f0a25df20f5951c5592fdfb13e5831b8f48b0ed",
   "videos/view0.mp4": "3444fb50046370afe9c3862bf16bd56a7e3552106588d87c3952ea5749d0ddec",
   "videos/view1.mp4": "e50ed03a1ccb26c5d53feb27c824c390406fafdd1d8cc4ab316eb95301de0647",
diff --git a/Simul-Robot/c001_episode_007170/source/cameras.json b/Simul-Robot/c001_episode_007170/source/cameras.json
index 9de5c9a763dbcfde3a8e568d1e2b285a84862549..7e967c4036e03dd88bbff8453e7da931619dbbf1 100644
--- a/Simul-Robot/c001_episode_007170/source/cameras.json
+++ b/Simul-Robot/c001_episode_007170/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007170/source/meta.json b/Simul-Robot/c001_episode_007170/source/meta.json
index eb09c0699c58687d9b2517c5064fa100e0584f99..ae71dcce096af5e8416694ff0077cf28e59e33bd 100644
--- a/Simul-Robot/c001_episode_007170/source/meta.json
+++ b/Simul-Robot/c001_episode_007170/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [17, 3, 1]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [3.16, 0.0, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [17, 3, 1] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [3.16, 0.0, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007170/source/visibility.json b/Simul-Robot/c001_episode_007170/source/visibility.json
index 1748f0006084046c39a3273101a9bd68e865e512..b2b978d75ee75e8c68851989f5cfddc3ebb2d980 100644
--- a/Simul-Robot/c001_episode_007170/source/visibility.json
+++ b/Simul-Robot/c001_episode_007170/source/visibility.json
@@ -48,7 +48,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "small_blue_cube": [
@@ -26629,8 +26629,8 @@
       "target": "small_green_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [17, 3, 1]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [3.16, 0.0, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [17, 3, 1] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [3.16, 0.0, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -47886,5 +47886,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007190/sample.json b/Simul-Robot/c001_episode_007190/sample.json
index a49a45a32390a03f5a9c8964c7aa748bc2a7aeb5..dca183df9c12365dd52bed83ca678d5e0bb7bc59 100644
--- a/Simul-Robot/c001_episode_007190/sample.json
+++ b/Simul-Robot/c001_episode_007190/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "88c9b9c5120d361492e08d5742c8a3416242e1d7e520af49da0ca427908a8294",
   "labels/objects.json": "19c268794ec351875bbdf7ec95f53b7b39e1c0e934aeb76181900bbe31d65979",
   "labels/visibility_states.json": "e666b931b472064feb625827ad152a2a6585aee055c6de13963f94a1b50f03fb",
-  "source/cameras.json": "55bfca24d788dc051deb06e847373c95933120573ce27e4b090cd2398d78ab49",
+  "source/cameras.json": "cc3dff5adab807e814ed1e3884691cb7278028005c0f0004dd598e25ab553cb3",
   "source/caption.txt": "bee92d0fd8c42798bf5e0f659f8e7b35621d75830ff1736ce096bf0dfa984560",
-  "source/meta.json": "f6a34ef60b7299aa2171871e6604730ca3666b4749fcc249340485ac473ec582",
-  "source/points_target.npz": "17025a984daadaf00def201b340b252d08454ab810efab2ed9f6ce2e97e3920b",
-  "source/visibility.json": "ba77b9a897cdde8b09d15ad2edc8fc68d59df864612e00bf381952c0cc3b1a81",
+  "source/meta.json": "f67ce089812f012932963c3d140d0085067f8788599672a46ea7178f0f36f22f",
+  "source/points_target.npz": "6f49399b78b52d7a83598e189afd524600bb1f57c6589619b6f17331c92dcc73",
+  "source/visibility.json": "fee31116f25ec8ff76e46a7eda4969e8f55be73df9cb7ad3d064ae26c5ee923d",
   "tracks3d.npz": "24a1ba8f3f896ba826d4e04799eb81af3e517e74df19f979fec3da69b6947938",
   "videos/view0.mp4": "0ad889c70afee3fd7e8ede684878e354017596b100ad6b029f0c60664a83395d",
   "videos/view1.mp4": "9662276294a529b98533af4f2054a0eedbec135e4566ad8091bfe80a9ce55e15",
diff --git a/Simul-Robot/c001_episode_007190/source/cameras.json b/Simul-Robot/c001_episode_007190/source/cameras.json
index b3c80009b73b1e838e7028cca893d79f7f1026bd..f96784974ef7e15bc5a1735800d025a7d91b1c7c 100644
--- a/Simul-Robot/c001_episode_007190/source/cameras.json
+++ b/Simul-Robot/c001_episode_007190/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007190/source/meta.json b/Simul-Robot/c001_episode_007190/source/meta.json
index 5316b49806925ca80f7c1abd86a19b98b9e26817..89238c869be5d27c8dd75fca84b2c04c7b83c685 100644
--- a/Simul-Robot/c001_episode_007190/source/meta.json
+++ b/Simul-Robot/c001_episode_007190/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [19, 26, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, 0.01, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [19, 26, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, 0.01, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007190/source/visibility.json b/Simul-Robot/c001_episode_007190/source/visibility.json
index 3a7046c3c195b7aaf7c58e1d566e1a8afdf8bd58..713d24265b92e855455eadf2094e14742d622572 100644
--- a/Simul-Robot/c001_episode_007190/source/visibility.json
+++ b/Simul-Robot/c001_episode_007190/source/visibility.json
@@ -48,7 +48,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_seasoning_shaker": [
@@ -30126,8 +30126,8 @@
       "target": "parmesan_cheese_canister"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [19, 26, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, 0.01, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [19, 26, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, 0.01, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -54183,5 +54183,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007250/sample.json b/Simul-Robot/c001_episode_007250/sample.json
index b022fbb0d0970803c1efc37a0342987bd55b80f8..10fe4bc4b1d0b373138b87eda5275d9e9b07dff2 100644
--- a/Simul-Robot/c001_episode_007250/sample.json
+++ b/Simul-Robot/c001_episode_007250/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "8ecf9d47d2a823a411f1c8f1af9457b3eb45a9b282e7f5efdac242ae9b8ed525",
   "labels/objects.json": "4aadde99e7fb8c4991866643dd66a8a9039d58f0c3ff73752092c0291e5f1b2b",
   "labels/visibility_states.json": "3ff3de22e84d48a5da17e50ad42b2e9024056b3ab4fdbb3f8abca8e6bf2ae9f3",
-  "source/cameras.json": "1334f82803a150c76e520fd7ce6bf5ad7061fab613709e6b7414f19c05d8255a",
+  "source/cameras.json": "ba540f2020155f8699bfc9ed60565db139a5539f2bcdfad7eaa4f7c2fe755046",
   "source/caption.txt": "aadf7c44a6e72a630581c9a659e1ba1e288b30867cee1c68a954b496ebe4abd0",
-  "source/meta.json": "f2d4d3106ff559b3d78369800f9493960edc69620df8ce005e937a88aea05645",
-  "source/points_target.npz": "8c7726ae922b3bba4cd28edef28edff6f5776c8e00c3eac2d4a8254f90f90ad0",
-  "source/visibility.json": "dc2d06bc1b3ba886ebaf6c271875bf785e33da9d36663fce194c9179d9cc925b",
+  "source/meta.json": "9631e1e851f72ea84a6e682a2f7c1cd76e44e9214785cbc639be0e56ecebbd63",
+  "source/points_target.npz": "8dba4e1a67cb66d7633cef3a6e94491bacbc6205b6a031d1bc3392fe1197e470",
+  "source/visibility.json": "1f77df2d8dbcbc4a6894f42bbced952c16c8012adb5c7a42caff2b96b4f973c9",
   "tracks3d.npz": "9e511e8d2b339dbb2d27cca0db50e24dd52fa46a96c6938e175208b082aefe9f",
   "videos/view0.mp4": "57aa801e729cfce6165c1033c9cab39316c16c5394a14f88ec6af9139b3903c1",
   "videos/view1.mp4": "c15ed7b3ef86d6333d02448608826d26dc0a413c8e8a25f5f24d1c46596e52bf",
diff --git a/Simul-Robot/c001_episode_007250/source/cameras.json b/Simul-Robot/c001_episode_007250/source/cameras.json
index fda5cbb62db41a6d29ba4206828860c30cc0edc4..7056745d82871a99e2a9816c58a48d25f36d6e1d 100644
--- a/Simul-Robot/c001_episode_007250/source/cameras.json
+++ b/Simul-Robot/c001_episode_007250/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007250/source/meta.json b/Simul-Robot/c001_episode_007250/source/meta.json
index c23d218bc1868924869a4a747af16d0e4995c01d..47585a9f5cbab1f363d65ee2557cf31021e4ad01 100644
--- a/Simul-Robot/c001_episode_007250/source/meta.json
+++ b/Simul-Robot/c001_episode_007250/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 2,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam02'] 의 최장 비가시 [18, 1]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.0, 0.14, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam02']: [18, 1] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.0, 0.14, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007250/source/visibility.json b/Simul-Robot/c001_episode_007250/source/visibility.json
index 232107234a54be178f102967bee3534b2549b9b0..567692744cb295227fded83cea2538dd8691b905 100644
--- a/Simul-Robot/c001_episode_007250/source/visibility.json
+++ b/Simul-Robot/c001_episode_007250/source/visibility.json
@@ -48,7 +48,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_frog_can": [
@@ -27329,8 +27329,8 @@
       "target": "yellow_energy_drink"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam02'] 의 최장 비가시 [18, 1]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.0, 0.14, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam02']: [18, 1] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.0, 0.14, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -49146,5 +49146,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007370/sample.json b/Simul-Robot/c001_episode_007370/sample.json
index fd65e5c7c30a1df5a7c402c87888e505fc678927..de57eea4d5ca0636b9189b1d977de11a07b7d9ca 100644
--- a/Simul-Robot/c001_episode_007370/sample.json
+++ b/Simul-Robot/c001_episode_007370/sample.json
@@ -220,11 +220,11 @@
   "depth/depth.zip": "e8d4546a28db2b76488a3986d31296c97608a88eafb7a45f84e67ba653bf9fdc",
   "labels/objects.json": "e670b49832dfe502275f41710ada470710e314056f96b195bfab151aca53e0ee",
   "labels/visibility_states.json": "570446392828983d9cbd533ce7cca79ae712803b651934c8bdf1af2f6bbec62a",
-  "source/cameras.json": "947e6e9ead3568950b512714dedfb50e7a2682e5ba5c1ba9f5ce818ecf869188",
+  "source/cameras.json": "a3e29f3bc5692470b96a0e29157e6e6dea597ec3888e69f0ea8da87f50d3f29d",
   "source/caption.txt": "bacf7ce7b8c291b30b7037e37c4fdd46018826a20b45c1ce0cb6d7dffa7118b1",
-  "source/meta.json": "1c0127043906098669c155626df8e0526666b73b30c83484972f4416d2adc75f",
-  "source/points_target.npz": "4e9cfb746c379416cc2663a7c73889626f7edf31ef11d108883ccc1e02fb883a",
-  "source/visibility.json": "126eeb620ae50ead86429e074c87374afe144812d37b6925581bd3560cc5bc87",
+  "source/meta.json": "7289e66c718555e2c841d0429df38759b4328e73fe5c35ee6b0f88d0754a8568",
+  "source/points_target.npz": "70f3743b881d424f1d0e7b47c2516b140228ee0e03a78758584bc0af1e31c201",
+  "source/visibility.json": "11ecb65e2fd3e152623c9734cd9ff9e5ffde0697c880e1ce045f7e19b8664c2b",
   "tracks3d.npz": "a43b5393110675cf4638ef39400c466bc6c28152a3330b907ab5f9b79d673d1a",
   "videos/view0.mp4": "4e9319bdefdcb48494ed7c4a5b0b0b1292fdd40517c73f09b3b39bd060a00bfc",
   "videos/view1.mp4": "b85e1dbf8326c0f7cafb447ddbba5ba5e77d6d3ad79c84ae4b7305603b1966d8",
diff --git a/Simul-Robot/c001_episode_007370/source/cameras.json b/Simul-Robot/c001_episode_007370/source/cameras.json
index 797bc8661b6c7cde3dec112255ce9bc8e71426b2..e24914a3a1d203c86a81c1d3bc9ed71c04232351 100644
--- a/Simul-Robot/c001_episode_007370/source/cameras.json
+++ b/Simul-Robot/c001_episode_007370/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007370/source/meta.json b/Simul-Robot/c001_episode_007370/source/meta.json
index c78e04f9626101a72499edaf986484caaef8d041..741cf10bd046d97d104a98ecb678abfc7e660a8c 100644
--- a/Simul-Robot/c001_episode_007370/source/meta.json
+++ b/Simul-Robot/c001_episode_007370/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 4, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 4f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 4, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 4 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007370/source/visibility.json b/Simul-Robot/c001_episode_007370/source/visibility.json
index 0de18d6d0f56f0c2a56b060115eff8afed55b2b7..34b5d9fa6147b6d001f661d5652fd14b5e6aadae 100644
--- a/Simul-Robot/c001_episode_007370/source/visibility.json
+++ b/Simul-Robot/c001_episode_007370/source/visibility.json
@@ -48,7 +48,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "small_green_cube": [
@@ -26974,9 +26974,9 @@
       "target": "small_green_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 4, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 4f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 4, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 4 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -48512,5 +48512,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007400/sample.json b/Simul-Robot/c001_episode_007400/sample.json
index b08ff6e74f36c574a16af7624b49cdf468af51fc..eb9c3484207cf4ccecfd159ef03220a33a2d39af 100644
--- a/Simul-Robot/c001_episode_007400/sample.json
+++ b/Simul-Robot/c001_episode_007400/sample.json
@@ -223,11 +223,11 @@
   "depth/depth.zip": "72e07d4b9202dbe87d392707172d74e1b093edac3318f5b55a3e83b2fe40c8c2",
   "labels/objects.json": "1c0b1e82e89024382ba548e787620f3231fa8d54f3d79eb5b8f12fb3883366f7",
   "labels/visibility_states.json": "4790f914e53e0c0dd92b09e82973d062688e00e5640c5e39a030afb2fbb4ab81",
-  "source/cameras.json": "d94b6911a46e0d8930597cca55990c6bdc55596f0d01e9b71d342de72ca4f39a",
+  "source/cameras.json": "6adc8f0fcdef349833358c56716916888d513259b6abff93381ffed8fc198b55",
   "source/caption.txt": "e5175d7d96c5c84a2a7f08e4fd606f404292ce77bc643535ba41c54c890d9164",
-  "source/meta.json": "c0784f1c357742a6220859b6585253a801b626c531e0a0a1034ef73fd6de6947",
-  "source/points_target.npz": "394e90042839965d5828cab6a203077593231d23c626cf8b1a264705e5b5c9b6",
-  "source/visibility.json": "e10fd375d443cb30e699069031777c3febf58042fd955ede6bb574019187d3d3",
+  "source/meta.json": "b7bdef7755436754f2873ca80ec9ee462a98da89deae093c80119ccfcf0288d6",
+  "source/points_target.npz": "957420c72cdaa51f1c2247fb993129b9372177b45d0b99aed81bc1512450382c",
+  "source/visibility.json": "c12528769bee859c5fd8fc92715711a03bb0b231e87a42bf559bcb1eaa77f070",
   "tracks3d.npz": "3be8e4daa4280f83f23fdf40c0a9c3ec4de57d4d156e1bb1fb1bbd72b7e41de9",
   "videos/view0.mp4": "f8b2e8231d812a7143afeb731e1a2d1dfd6ee8be9fb82a6ae9e08dbdfef29d97",
   "videos/view1.mp4": "8743927b10d46d0072f4c327ab8d037ce95cd468a6d1e1cdef654493e6812652",
diff --git a/Simul-Robot/c001_episode_007400/source/cameras.json b/Simul-Robot/c001_episode_007400/source/cameras.json
index b78bb656e19eec564e16f113405391f93cc455eb..c9c91843a59d335d081f824c138bde41022e280e 100644
--- a/Simul-Robot/c001_episode_007400/source/cameras.json
+++ b/Simul-Robot/c001_episode_007400/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007400/source/meta.json b/Simul-Robot/c001_episode_007400/source/meta.json
index 3d1fc9f434b0afc08106e95dce2e82a3e27c0c74..85905b93d0d94d7e86419d44f0f0f49a1e68a57c 100644
--- a/Simul-Robot/c001_episode_007400/source/meta.json
+++ b/Simul-Robot/c001_episode_007400/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 3,
     "hard_reasons": [
-      "①캠 ['cam00'] 의 최장 비가시 [16]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "②전 캠 무가시 연속 3f > 0f — 구간 [[615, 617]] (최장 (615,617)). 정보가 없는 프레임은 **불가능**이지 어려움이 아니다. 합집합 가시율 0.9954",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.14, 2.01] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00']: [16] f < 30 f (a single camera sees the target over the whole clip)",
+      "② consecutive frames invisible in all cameras: 3 f > 0 f — runs [[615, 617]] (longest (615,617)). Frames without any information are **impossible**, not hard. Union visibility rate 0.9954",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [0.14, 2.01] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007400/source/visibility.json b/Simul-Robot/c001_episode_007400/source/visibility.json
index 8caefae04117e530ff56ae2849fa299ab7c048c9..1146d201812efc075b7204df6c1dbba5029d31d3 100644
--- a/Simul-Robot/c001_episode_007400/source/visibility.json
+++ b/Simul-Robot/c001_episode_007400/source/visibility.json
@@ -48,7 +48,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_frog_can": [
@@ -32635,9 +32635,9 @@
       "target": "small_red_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00'] 의 최장 비가시 [16]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "②전 캠 무가시 연속 3f > 0f — 구간 [[615, 617]] (최장 (615,617)). 정보가 없는 프레임은 **불가능**이지 어려움이 아니다. 합집합 가시율 0.9954",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.14, 2.01] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00']: [16] f < 30 f (a single camera sees the target over the whole clip)",
+      "② consecutive frames invisible in all cameras: 3 f > 0 f — runs [[615, 617]] (longest (615,617)). Frames without any information are **impossible**, not hard. Union visibility rate 0.9954",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [0.14, 2.01] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -58693,5 +58693,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007420/sample.json b/Simul-Robot/c001_episode_007420/sample.json
index cfb3ab02791e655823cbf619927914478ff92b53..54fcb4731678da64c36508c4d8b24284aefbb53e 100644
--- a/Simul-Robot/c001_episode_007420/sample.json
+++ b/Simul-Robot/c001_episode_007420/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "7686d95317b06ddce73c478f380215938839735a423b576d564f67fe3147f0c3",
   "labels/objects.json": "c96c1e48af5a65ec92fed4a1645a597641226d93f99c1debbb46efecf13fb66e",
   "labels/visibility_states.json": "34f15943d82093d2041ed174a09dcbdf5c3e574e88e4d0e6027670af5834e7e6",
-  "source/cameras.json": "44e7035e8d72147bc93d7ba6ff5c15b94d03860b8958bea877de8d5069d03fb2",
+  "source/cameras.json": "daf6ef17c8d16d964263d04e1c435f6b4c833477e414a6069bd57fc67a574624",
   "source/caption.txt": "df5209ef8eb40535f2f82eb1c535271ce19568996b455a1c50e7abe65cf7fc40",
-  "source/meta.json": "48fdedee020709fcbf7e7f5eca124b928309659d425a6154b175da813fc285f1",
-  "source/points_target.npz": "950145334aacb83106f40c3060af3959e5d3e7f6a60c01a33d5c4414dd2d7447",
-  "source/visibility.json": "1f5b61628948d228d51b5b874fb79138e7d25f3e1e076ac6af5882a17c1b179c",
+  "source/meta.json": "346b4ae1a7afa7c13ab28b15896231bb665dd39e17a6ad5bc1fbc5d0de299a60",
+  "source/points_target.npz": "f87c3d531799627b7e479482159c67fe4036246c7978ffa53196aff0c24e3f74",
+  "source/visibility.json": "321823573564d94298917887a13a179f5e92f5e2bdbf5cddc39857c00b7accf1",
   "tracks3d.npz": "e4edaaf36ddb9ebc2201a56d57ee1ebc3824ac8ccaf9dc94f163dcf81b8148ca",
   "videos/view0.mp4": "f2a4dacd8fa14116fc6a4df4098317e27a208f4516840a84cbaeee50934d0ae6",
   "videos/view1.mp4": "194c90ef1889e0bae2cf7abc88f3a47cfadba3add7dd2dc4c124122b71fd0515",
diff --git a/Simul-Robot/c001_episode_007420/source/cameras.json b/Simul-Robot/c001_episode_007420/source/cameras.json
index 3f995202c853ab78ae00b5a66b2ab074b1c50fbc..3222604553995377fe5975b940520904c539267a 100644
--- a/Simul-Robot/c001_episode_007420/source/cameras.json
+++ b/Simul-Robot/c001_episode_007420/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007420/source/meta.json b/Simul-Robot/c001_episode_007420/source/meta.json
index 6caa3b68ce87e945f19fd88c14117f050d030624..edb25f1ad96d0cefb55b11aee61e4bbf82dfa1fd 100644
--- a/Simul-Robot/c001_episode_007420/source/meta.json
+++ b/Simul-Robot/c001_episode_007420/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 5]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 3.31] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 5f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 5] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 3.31] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 5 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007420/source/visibility.json b/Simul-Robot/c001_episode_007420/source/visibility.json
index 5c4fd2eb8e851529d37e8cf324a9656bce66fa55..f0132be357a163befb8ff00d1a873df324f32681 100644
--- a/Simul-Robot/c001_episode_007420/source/visibility.json
+++ b/Simul-Robot/c001_episode_007420/source/visibility.json
@@ -48,7 +48,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_seasoning_shaker": [
@@ -25873,9 +25873,9 @@
       "target": "sweet_corn_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 5]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, 3.31] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 5f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 5] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, 3.31] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 5 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -46531,5 +46531,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007490/sample.json b/Simul-Robot/c001_episode_007490/sample.json
index a6ef3dc94acd5908d894c78bdf5ff63a99519d5e..344bc9557eca18e23ec89e6a5cebec02f322b3ad 100644
--- a/Simul-Robot/c001_episode_007490/sample.json
+++ b/Simul-Robot/c001_episode_007490/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "24050819916c12da3ce7ee1ee40d69a19ed764373e312393c5d97b87d6eacf06",
   "labels/objects.json": "c0bfd162c70f142c8b6707cae6b0cdf0be311d74cb4f3711181fca771fcddefc",
   "labels/visibility_states.json": "ee8a61995ac4971ca83b5527f87dfdc1b2123d76e9252a392cc82b6381cbc49b",
-  "source/cameras.json": "a9b9849abdf7de36c49e7251e3eb5c4aa10e0579bc9734ce3f93c07adca0d5ba",
+  "source/cameras.json": "4555782718dfd1667f38ef846326ca3e653ea4e15631b604623ff1b7710e8bca",
   "source/caption.txt": "710d60ae476378623f0702fe74ae9d98e09a334ab4235e87d81ed984f15dcb9f",
-  "source/meta.json": "5402aeb77cd0f469ef8c5dd5c80562c448bf078c84e725d0dfc92bcad82cd118",
-  "source/points_target.npz": "9b28784642b85482249b14b53ba22e520d7c344c3e3fb69ec0abb1b9b0cf3d93",
-  "source/visibility.json": "34b7c519c7d434aeb4995ce485c5149a039a92a4fe67e6e016b5303d86594093",
+  "source/meta.json": "57ccc16b993fdb70de5a42f6db2b2d69a476578b85931e6e093aef7fb37df9ad",
+  "source/points_target.npz": "d958bfade79c6de57fc1924471c9a12c6486d48d50760e6ed0e73dc315c0ec6d",
+  "source/visibility.json": "d15e0aa6a495933a2fa40fb33f71f5b74a7bfcfaef22190be3f145764e4ded26",
   "tracks3d.npz": "04d03e53662d3e1682df75cd5770019b91f95f10e5fcb3e9a00dd58ac7270848",
   "videos/view0.mp4": "c1c6578474854bba0b9cf042b3f515bbc2ed5703e0df9d57a3503a5000df4ff5",
   "videos/view1.mp4": "4dee0d946640d4d0021bdf68f2737dd8eec0261ecc0fa042d4be9efc39c6a148",
diff --git a/Simul-Robot/c001_episode_007490/source/cameras.json b/Simul-Robot/c001_episode_007490/source/cameras.json
index 366bc6743b33a3a4980fbb503df347d42fb901b8..d5210c33d7da29291903a49d36decf05e3ec43f2 100644
--- a/Simul-Robot/c001_episode_007490/source/cameras.json
+++ b/Simul-Robot/c001_episode_007490/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007490/source/meta.json b/Simul-Robot/c001_episode_007490/source/meta.json
index 9415880eff050f725e97a70ce52dbb6fbd2e5d17..87c70a0e1e286886bad2b28f390dc04acb640ee4 100644
--- a/Simul-Robot/c001_episode_007490/source/meta.json
+++ b/Simul-Robot/c001_episode_007490/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007490/source/visibility.json b/Simul-Robot/c001_episode_007490/source/visibility.json
index ecdfe5171b254f05636364945b984160307bdf15..66426f4868c1429bbbcf0d4fadf30cadf06db4c0 100644
--- a/Simul-Robot/c001_episode_007490/source/visibility.json
+++ b/Simul-Robot/c001_episode_007490/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "black_energy_can": [
@@ -7318,9 +7318,9 @@
       "target": "painted_tin_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -13136,5 +13136,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007500/sample.json b/Simul-Robot/c001_episode_007500/sample.json
index 4a00afe7a5bd3ba0c61cb2d82e23114584debb93..2c2fd2005119e69b276bb5b42c8327f339829e57 100644
--- a/Simul-Robot/c001_episode_007500/sample.json
+++ b/Simul-Robot/c001_episode_007500/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "019bd88bdd5c9c0a5395a25aaed4fa6e6e74792d14c4d9004e8ae9a6ab836c4b",
   "labels/objects.json": "d90f2073ea47d6a83d8b5ab159b71d6320f09fa472cf3202304602ea32cbb2de",
   "labels/visibility_states.json": "88404e8e05b319bda2de8a7a4067da29b8110cfdd975cfdc37c5193b37f5964e",
-  "source/cameras.json": "52d3d693d5ab96062cbfc595d73d0b6b4dcdc38d61663f3a2f1c0b8cc2cf3be6",
+  "source/cameras.json": "f8c23a5bdaddde418368fdcf383097d678de6735cb231f696a2ff7012715176d",
   "source/caption.txt": "3a90cf69570d63612d0dddb5dfb8f1e5eb07b5f5c71edc79269d897bde9a53e7",
-  "source/meta.json": "d2b619cdeb3489da84e8d635d71e480aba97ffa74842ac2655beb95e1206bca1",
-  "source/points_target.npz": "2b0011f12fc7c19573a288055a1fd93519a919d3f4bf1b69689899216dd9f63e",
-  "source/visibility.json": "81667cc08654bad94cf9173922bc9fc9171907d8cc2789677e26df969f0c5397",
+  "source/meta.json": "c7975140ff241b1ed67c39c26912a79188b31f29cb3efe5a11eccd0bc569d166",
+  "source/points_target.npz": "7339755b8cf5b328c75fb3f61386242ca83c0dee1dc6270219db786d15337b77",
+  "source/visibility.json": "59852b3ae972b209f972c5742aa67c4f6b52b44d5c356ac569b920492466cba9",
   "tracks3d.npz": "b4b072500630c73b9ce0d07db1dd6b41e8c321c6375069a0c42691f2474efd4c",
   "videos/view0.mp4": "3f504c78e58fe7a13ed25f53a385cc7a2b29a79f0670f6a99813bb376d334b82",
   "videos/view1.mp4": "569ca6ea59636478585580424059993ff999e8cf38c9cb56ca3470d9ac5624d6",
diff --git a/Simul-Robot/c001_episode_007500/source/cameras.json b/Simul-Robot/c001_episode_007500/source/cameras.json
index 1b4ac45e1e891d226180d9a22233807336a4bd5a..4d54973b4e679939d13ff67951ea5fb793873983 100644
--- a/Simul-Robot/c001_episode_007500/source/cameras.json
+++ b/Simul-Robot/c001_episode_007500/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007500/source/meta.json b/Simul-Robot/c001_episode_007500/source/meta.json
index 55e700da1856a43fc7e9ee57b06677fc1ec304b0..f9e9cc29092e5ae99bb284ca218d0ca9e5cf1e3e 100644
--- a/Simul-Robot/c001_episode_007500/source/meta.json
+++ b/Simul-Robot/c001_episode_007500/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007500/source/visibility.json b/Simul-Robot/c001_episode_007500/source/visibility.json
index 19ec6a4cca5600e3cff31f01356519e5ae69d013..4ce6f11797f015db80b92cbf9ede4b0d206906bd 100644
--- a/Simul-Robot/c001_episode_007500/source/visibility.json
+++ b/Simul-Robot/c001_episode_007500/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "magenta_plastic_crate": [
@@ -6168,9 +6168,9 @@
       "target": "red_energy_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -11066,5 +11066,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007590/sample.json b/Simul-Robot/c001_episode_007590/sample.json
index 4a88bce3bf1bca292d70ae50a90815eb11f57c2e..36bcf5d595b350e3f80975adf48e95583da3e6f3 100644
--- a/Simul-Robot/c001_episode_007590/sample.json
+++ b/Simul-Robot/c001_episode_007590/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "e63b992132b33cc248a2f4e5d6d0071a21b809ccd5abe217050e20db3e932f9c",
   "labels/objects.json": "5bb770d30029a127a16d20a59b8eacedb96f7f99082666ca76d46918af4574a1",
   "labels/visibility_states.json": "e6bb39e471c7450e4504efd28c7bd711d96f5375f4498e29105005b6cfd8188d",
-  "source/cameras.json": "595e10919d6b2ef30dbc96f183cfc1523ae85d52b9be8c45b72e78e6d5970a80",
+  "source/cameras.json": "4f1191adf2f1c78b94deb0d163fa6d302e2a694294ae493ca8f4eaaab32d50dc",
   "source/caption.txt": "a2f83ff206945c96dc2a12d76fa53ba004c2fb250ebfed35bc7c9768822f1620",
-  "source/meta.json": "776c257d4b30a264284c2d5dc56d140c3ef6eb6fb74f49430b0c421516019711",
-  "source/points_target.npz": "eaa2fd2d900125b39ba6c20be32259b274d826eadc43291e53a348c83a715585",
-  "source/visibility.json": "ebe416fce7973d0559ae1ea8795a2eb05a339576d47cfc3656ec07dcb96d0cbf",
+  "source/meta.json": "1c56fbd16e95db3f836fb3cef337447b4ed3dca297b307f5aacbe2dfe8bb88e8",
+  "source/points_target.npz": "fa46fd74704212f56ce7f88e0d151cec20b989e099fe3371bf9b3fc8d6ea5c89",
+  "source/visibility.json": "f772bce51eea2f530dbf6435a740abbc86566ce01f3bf034d9aac6fd224459cd",
   "tracks3d.npz": "101d8ca3eeba448cf12f7565d4061ee6d4ff692092aef5ce0565655d201e9372",
   "videos/view0.mp4": "7fedba6a41d6dc9a3d33d8c79e81df9b1cec39b207c29eb9cd3649801abd36ff",
   "videos/view1.mp4": "21c1fac8e7418460cf05971e8b7ff949f898d16433efd389be7870a5cfb80fa6",
diff --git a/Simul-Robot/c001_episode_007590/source/cameras.json b/Simul-Robot/c001_episode_007590/source/cameras.json
index 6ad1612a5600cd91044641f518aa713c9bc1e6c4..95bb59a32891c84b3d776cd96aa654dad1d672b4 100644
--- a/Simul-Robot/c001_episode_007590/source/cameras.json
+++ b/Simul-Robot/c001_episode_007590/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007590/source/meta.json b/Simul-Robot/c001_episode_007590/source/meta.json
index 22cfe8b02d2c5fb2a6f1231629abbd992563ddf6..88e251dadaf5bef4d936b7f9e27b7357d7f8e788 100644
--- a/Simul-Robot/c001_episode_007590/source/meta.json
+++ b/Simul-Robot/c001_episode_007590/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [1, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 1f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [1, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 1 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007590/source/visibility.json b/Simul-Robot/c001_episode_007590/source/visibility.json
index b311d4b897374a9e19bd82306210c34af3b75784..08aa4158bc82b73b99ac56dfe3cb25daf3310375 100644
--- a/Simul-Robot/c001_episode_007590/source/visibility.json
+++ b/Simul-Robot/c001_episode_007590/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "parmesan_cheese_canister": [
@@ -7321,9 +7321,9 @@
       "target": "small_yellow_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [1, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 1f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [1, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 1 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -13139,5 +13139,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007600/sample.json b/Simul-Robot/c001_episode_007600/sample.json
index c2cdc1eacc985fab62f66c71ac10ccc53c763883..ad1f6fbfbd61322a6735cdc174fcaade02a3b414 100644
--- a/Simul-Robot/c001_episode_007600/sample.json
+++ b/Simul-Robot/c001_episode_007600/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "c0e3184e14de9962beb6d6bf4b2e25443f85e51dc2766035e67f22503fd5c718",
   "labels/objects.json": "7317cce762a52cfc3282e7ea5cb5e847c15b58f94bf4d9c42fe3c0a6e1793c2e",
   "labels/visibility_states.json": "21aabdef3cfff7aa244bdf0394d1ac908649704113d0319e231a7b8945c6a85c",
-  "source/cameras.json": "ee528450b360e97f64baa745d750bb726d953c5170374df204a9d1f243c48667",
+  "source/cameras.json": "65d2146a976b74aba80a73de1bb080192da2cc917be7141e5e441d6e658001cf",
   "source/caption.txt": "430c352921339bb55b46bdea2aee31094037b05b300b87fdbc189da99642a03f",
-  "source/meta.json": "52cb1c53fce45bb60606c0c161b329098d1c4126b82216046a1b7a6dc64221aa",
-  "source/points_target.npz": "e79ddb8e0e658205fffe2de55fa7d4858f56f716bf7648eb6f6973db2a19a65c",
-  "source/visibility.json": "d1c17631cc1baef544e572deaf802b32f870df41e77057169689da04130e3a0b",
+  "source/meta.json": "7fcd30b57afbe79a8b536f6ebcda3361f03295b82184f046f443583802f8cabe",
+  "source/points_target.npz": "3a9e210834afbad9c5cd1318829fc43c3c8c5688b5adad9ecfe585ab2c5fcacb",
+  "source/visibility.json": "d4aaa7d2a819f72e0ef5850312ccdd09f03f8c837d302d3126f8e8d7f903bed4",
   "tracks3d.npz": "308779af631a809a30080a2665f8c5b6e486d38f077f52618385a29ec14c1646",
   "videos/view0.mp4": "d5f1978e33d8d0b65fe21b2c7ea6f98450153221be658b9750be6f44c9611546",
   "videos/view1.mp4": "490439e75dfa70dcf188c94f56163b07069121368edd9236973d6c9ce029224a",
diff --git a/Simul-Robot/c001_episode_007600/source/cameras.json b/Simul-Robot/c001_episode_007600/source/cameras.json
index 723362ad14cabab1f1e2fdc19e4947612c205642..ce8f3a8de16fab95102a34e34a9f9f6f2c82ed16 100644
--- a/Simul-Robot/c001_episode_007600/source/cameras.json
+++ b/Simul-Robot/c001_episode_007600/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007600/source/meta.json b/Simul-Robot/c001_episode_007600/source/meta.json
index ccd95c27097507d4c8e1e632388bdd5de95e7bed..d6e76f27efd81384c996c27bfa4e309d237642bc 100644
--- a/Simul-Robot/c001_episode_007600/source/meta.json
+++ b/Simul-Robot/c001_episode_007600/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 2, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, 0.58, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 2f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 2, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, 0.58, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 2 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007600/source/visibility.json b/Simul-Robot/c001_episode_007600/source/visibility.json
index 08459b2294f0bb52142834a5811e5ae73666366c..d40ab14bcd6b27356fad474c5fa6e40409bbc67c 100644
--- a/Simul-Robot/c001_episode_007600/source/visibility.json
+++ b/Simul-Robot/c001_episode_007600/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_plastic_tote": [
@@ -6171,9 +6171,9 @@
       "target": "small_green_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 2, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, 0.58, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 2f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 2, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, 0.58, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 2 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -11069,5 +11069,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007840/sample.json b/Simul-Robot/c001_episode_007840/sample.json
index a8eb176b0557c706e150559814bc3dee6d638985..5e5d6a4c5aa95a5f379f1cfe383655f9b6a5eb15 100644
--- a/Simul-Robot/c001_episode_007840/sample.json
+++ b/Simul-Robot/c001_episode_007840/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "939acce4a04d39937b92759771ea01ba2ede90b604b08707a902d5bdd70176b0",
   "labels/objects.json": "764693df49760f864ee5f36f07dc3dfd86bbd6607ced626d4c225044d12dd003",
   "labels/visibility_states.json": "120323c681231fe25a583088a4940e527adb61d01669e98aeb6dc2e28c00f5be",
-  "source/cameras.json": "ae1777024cef39bac7930fddb7d9999d750105cf2ce9c148ff4a16b2d1f9b5a0",
+  "source/cameras.json": "a1ec20028bc03dada1ed03ee75b55c325fae4fb76782a38e1d96b71b8600da52",
   "source/caption.txt": "6f83c18fd19a5788f6d18e266e301273991889785473ae4c976364efcd98db7d",
-  "source/meta.json": "2add2406fc01f311bd7c362096e8a2a6eef0d1a4b1cc6053408f6b93b12555c2",
-  "source/points_target.npz": "9261c1bdf2c3e2b2d4c5154fd6633e3dfe843119fc6c46f0f9a9f836e5f9d3eb",
-  "source/visibility.json": "0bab1f41eb485779af63fe3963e6d29f520b7ef0f88d504a0909a095e43f6f26",
+  "source/meta.json": "f1a037ef420b81fe9c292b6f8973fd4d9c4de73b158b6bb5864cf248d6c37981",
+  "source/points_target.npz": "992722861bf483e30db3a2544d8a7dc3c3ab85ce730cdd9e18317b80ff2a8628",
+  "source/visibility.json": "b5e48acf4badec027d4b690dbe63f06c7b5262a4dd7fe67121efb87357ab36b4",
   "tracks3d.npz": "c3506d11372483e9c96027249e96185479ff5e1cd06aabb8e0da46d047bca694",
   "videos/view0.mp4": "fcbf2e8bd23db3185822c1e6887851bc14ff89e93734155cbec292bdc6ee76ee",
   "videos/view1.mp4": "28865453227c2c1a85de0bb64ae3fa6be4d0c009af1ef7155a15414baab5c8bd",
diff --git a/Simul-Robot/c001_episode_007840/source/cameras.json b/Simul-Robot/c001_episode_007840/source/cameras.json
index beeacbab9c5cc548567f257ba08dd74143a8daa8..c83719eabeac6c3d10fbf9333e5588b5fc3ef460 100644
--- a/Simul-Robot/c001_episode_007840/source/cameras.json
+++ b/Simul-Robot/c001_episode_007840/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007840/source/meta.json b/Simul-Robot/c001_episode_007840/source/meta.json
index 5b7faffcb0e51b590f472ff494cf4c56fe47d853..ccaa490d574cc818a86e4b0c791c5cb6ded0072b 100644
--- a/Simul-Robot/c001_episode_007840/source/meta.json
+++ b/Simul-Robot/c001_episode_007840/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007840/source/visibility.json b/Simul-Robot/c001_episode_007840/source/visibility.json
index 758105d1a1ec2d13eef6c3bc0e5d90b0c2cb48a0..2db6519e32d129bb593f136d0148511cda6c99c7 100644
--- a/Simul-Robot/c001_episode_007840/source/visibility.json
+++ b/Simul-Robot/c001_episode_007840/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_seasoning_shaker": [
@@ -4168,9 +4168,9 @@
       "target": "yellow_dog_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -7466,5 +7466,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_007890/sample.json b/Simul-Robot/c001_episode_007890/sample.json
index 69602c4467c9c2290b3f68d583966842b48b21aa..94046a6ca74de47348156581e7512100f29f87c4 100644
--- a/Simul-Robot/c001_episode_007890/sample.json
+++ b/Simul-Robot/c001_episode_007890/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "ef7dea3e9cd9e0204d524908c159ea472be00724e556741037c5fc0b62c720ba",
   "labels/objects.json": "cad334d5438c17a48b071628db95d79b704f0a66d10f225893806e9286f2a932",
   "labels/visibility_states.json": "7ea89ff433fe9fdced877454a33dc517fb3572911e262d8cb92094f1962fe3dd",
-  "source/cameras.json": "f380ce2c2108e9c5ac2a1dc5fa798c7b63db67ad9964fb18ecdb92b52f1aaf54",
+  "source/cameras.json": "11b31ca21a014c9b1af30e9447ffb5830902198c47ff2585e0a49de4187f9168",
   "source/caption.txt": "479912b81f86822b2de724521071976b4847547494eb17f2495cdd5d33996042",
-  "source/meta.json": "030430dfd0df7611be6590a6d62627f74773828a2940f525aea50619085440fe",
-  "source/points_target.npz": "168270fb7fbe0d5cee4b4ece1dae65e10601d599b1198e33be16d57317e18fe1",
-  "source/visibility.json": "a07c20065157d8cd94802082e835f134c78bd6f77081973319131770ba6dc700",
+  "source/meta.json": "b26cf0cc0a5ba3679233183fb5c9483030d2429e3651074f884f26979322beb2",
+  "source/points_target.npz": "ff748b6a6acdb1464b2bc6ec639c3b697791942df84e9d521659b250de4c19ef",
+  "source/visibility.json": "3021009500c3d203190cf34504f38d28522bd3c6f8a3e6b6098df06fda392a6b",
   "tracks3d.npz": "7c77a918cde761c5613f65076dff4174b9ae6c77eb8e28b0bea5ee785ae0b63f",
   "videos/view0.mp4": "b72e8effce2959009338898a793aa033805965be3bc1b74056f54a3bc9286ff7",
   "videos/view1.mp4": "19a502b5e07bc8833fdd925a4b6f2871fdba5046cffcdb89c3cdebdd62cc3ab4",
diff --git a/Simul-Robot/c001_episode_007890/source/cameras.json b/Simul-Robot/c001_episode_007890/source/cameras.json
index eba041c788cf4b6576e25fa30410bc24cc2aa4e5..c62e994484102929b8903e058dc6e890527e58ad 100644
--- a/Simul-Robot/c001_episode_007890/source/cameras.json
+++ b/Simul-Robot/c001_episode_007890/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_007890/source/meta.json b/Simul-Robot/c001_episode_007890/source/meta.json
index e9ae915c063610f9a6b17420b9717e1223ba9124..6f53053e2dc0e98cef47026e9cf9370aa9505539 100644
--- a/Simul-Robot/c001_episode_007890/source/meta.json
+++ b/Simul-Robot/c001_episode_007890/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [4, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 4f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [4, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 4 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007890/source/visibility.json b/Simul-Robot/c001_episode_007890/source/visibility.json
index 7850b511d7759c60a76a65bcd0d3799dad21dc62..1c7922d364f0209f036a407f11afece5e522bd5e 100644
--- a/Simul-Robot/c001_episode_007890/source/visibility.json
+++ b/Simul-Robot/c001_episode_007890/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "green_energy_drink": [
@@ -5121,9 +5121,9 @@
       "target": "painted_tin_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [4, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 4f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [4, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.01, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 4 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9179,5 +9179,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_008000/sample.json b/Simul-Robot/c001_episode_008000/sample.json
index 87b529094daa31e52158feb172c389c825000960..1c89c2ba6f168f5c3207a7d804312b0e10b8ea83 100644
--- a/Simul-Robot/c001_episode_008000/sample.json
+++ b/Simul-Robot/c001_episode_008000/sample.json
@@ -221,11 +221,11 @@
   "depth/depth.zip": "39fd609f1efbc98ea396af14f3570687c76e110e946bd6424094e609b8b07387",
   "labels/objects.json": "b8d49bf10a12a1d1f21dc4cb04539e4156fd5bea28b79586edbdb556d23b3e4d",
   "labels/visibility_states.json": "ec466cb2752cdc36dffcddd018a36cb9b0b24bd1bfb170825b9b7bc10a26221e",
-  "source/cameras.json": "30ff2f77188bf851cba8d04f8bc11eec104c37c80042d317cca7a5cd9e5b590c",
+  "source/cameras.json": "90714eb126cf7bef12d6d84bb1d2060742372bb77a552d44d1859a4eb2948bc5",
   "source/caption.txt": "e6b3c33f20b6c3b52922afe627838ddb7882395e2337765b19adc364a04f7f36",
-  "source/meta.json": "235104613e628702678c086680e90fc932a6f26f4445f8527a29290d3425f00d",
-  "source/points_target.npz": "444977cf510030db38110138237bf8c9ad6082724b63788e6aac01aa1e05e9eb",
-  "source/visibility.json": "525e00b49b7120a2e01a7fb789c937faf29f8509b6fdf92db5c771f06dad266e",
+  "source/meta.json": "a4efb51faeda02b028d05da2b08af7a7c27872273f5a451614cd380def297a27",
+  "source/points_target.npz": "59731cf5bc362c4cdeba8cd6da1e16444e097136530dd4a3a399238c3dafc4a8",
+  "source/visibility.json": "c761fda8459b6de20426c1d2428a78ecc1c4e811d6cade721dcf911ccb7302f0",
   "tracks3d.npz": "b68db81b6d7840d92fb7b3ee7e48a23f7d66a53edc4486910f4db611caf66384",
   "videos/view0.mp4": "ebde9f266ecafc85cdb48e20fcc7cf3d92c5bf5e796c1712025119ae2a8aa8d7",
   "videos/view1.mp4": "5b31876e8316fd74f57361ad82ca4534dd28c571bce2a6be3f18a537e38792d9",
diff --git a/Simul-Robot/c001_episode_008000/source/cameras.json b/Simul-Robot/c001_episode_008000/source/cameras.json
index f1425a532cd99293d59931359f8e412faba736c6..e6dddbc89681bcfa218543f28e3fb695f9243a7b 100644
--- a/Simul-Robot/c001_episode_008000/source/cameras.json
+++ b/Simul-Robot/c001_episode_008000/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_008000/source/meta.json b/Simul-Robot/c001_episode_008000/source/meta.json
index bde022762512508103968508984d43c2b4a2b702..d2b9784176b1e2b5c5782fcba9aaa73eb47e4ce1 100644
--- a/Simul-Robot/c001_episode_008000/source/meta.json
+++ b/Simul-Robot/c001_episode_008000/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01']: [None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_008000/source/visibility.json b/Simul-Robot/c001_episode_008000/source/visibility.json
index 26398ec0858ed348291c5c9e8f4071b8557a7e63..bf9c97bce0eb67b76039569f195b378d3d87b5a1 100644
--- a/Simul-Robot/c001_episode_008000/source/visibility.json
+++ b/Simul-Robot/c001_episode_008000/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "magenta_plastic_crate": [
@@ -6171,8 +6171,8 @@
       "target": "small_white_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01']: [None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -11068,5 +11068,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_008050/sample.json b/Simul-Robot/c001_episode_008050/sample.json
index e411d146ffe4a7d07481271484e5064b9373524a..592543a46b86c779ad06233da4a1bca686d52f1f 100644
--- a/Simul-Robot/c001_episode_008050/sample.json
+++ b/Simul-Robot/c001_episode_008050/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "91b6b28d8df4c6b83984f0d433247968e40fc2948c4dab6957a756d6f11aa7fb",
   "labels/objects.json": "e0e06a6ef8e896c65d0fcb3df636eab5ca3e8575d0bebc74168590c29990cc86",
   "labels/visibility_states.json": "05e55e717d0e1c172c39515dc4ef7c3137eecf9543bb9ab676b9d85ec27fbe7e",
-  "source/cameras.json": "0b404d6c9fe039ae80ec80c1c5615a3045311777838ca1fa9973b970dfc73699",
+  "source/cameras.json": "0ee3788e57b8dac2cea6c2cd148c5563087e5f7e08b0b5b3c441813fd3bdc924",
   "source/caption.txt": "d36bca6b32892e489389eb7f03aa0d2b800261d27d031a7102f370cbe7e6fce4",
-  "source/meta.json": "5190d70045a499050ec99bd1e29d19f2eeb5b40156dfe71efcfaeaa0fe8d557b",
-  "source/points_target.npz": "f7870a9e7e8558940ca39c32fb50059ee6b1db8a5e70b961b41500b0f292f8ec",
-  "source/visibility.json": "e0b27758d16fcb2c1e1306056520c86cb571a4d2df6ef6f050f14f1822416310",
+  "source/meta.json": "a734f6b22c08191efa0777a05abd63626e1362ac4c498ac4f6139a8ae24c4894",
+  "source/points_target.npz": "e34460885baf2f93896d3b65f8a44a632fb4f7dfe40c0eaf38934452c4ba959e",
+  "source/visibility.json": "20f3cdaf96a44c7a16c1d84ca6bc501e8ca640dba0cb7e6ef89f2bef87f76a10",
   "tracks3d.npz": "af641fd7f17f3a99a6e758f5e9a1e4589530d6a0d41b050389f55a8c04951c30",
   "videos/view0.mp4": "0c7c34412507c3623a4601ac0aec8d88614aa18080f773d869f1a0dcb7272c35",
   "videos/view1.mp4": "3e645c5266547463d14b69492004dd1fb3521fdd03862523616151ac79866ea9",
diff --git a/Simul-Robot/c001_episode_008050/source/cameras.json b/Simul-Robot/c001_episode_008050/source/cameras.json
index b5b00b8c587e78a60c8a404e9afff04ef24905ae..4913fa002824aa38b95beedc46d6cd084d130252 100644
--- a/Simul-Robot/c001_episode_008050/source/cameras.json
+++ b/Simul-Robot/c001_episode_008050/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_008050/source/meta.json b/Simul-Robot/c001_episode_008050/source/meta.json
index fffc245f6604501637cea9f0ab790470ceebb8d9..e8da6f5ca0fd7d99c037312038dd879480c6863f 100644
--- a/Simul-Robot/c001_episode_008050/source/meta.json
+++ b/Simul-Robot/c001_episode_008050/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_008050/source/visibility.json b/Simul-Robot/c001_episode_008050/source/visibility.json
index 4df8f9dbaaf6fdab6c40160fca2783ff0a1195c8..6dbd88f01578a429ba8b540a721586115ef2db56 100644
--- a/Simul-Robot/c001_episode_008050/source/visibility.json
+++ b/Simul-Robot/c001_episode_008050/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_seasoning_shaker": [
@@ -7318,9 +7318,9 @@
       "target": "yellow_energy_drink"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -13136,5 +13136,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_008070/sample.json b/Simul-Robot/c001_episode_008070/sample.json
index e672be581d57bef3d5c6ccaebc69c7c57a494125..6da9957052fbd25b0d89c2022240acc9a4235c7a 100644
--- a/Simul-Robot/c001_episode_008070/sample.json
+++ b/Simul-Robot/c001_episode_008070/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "cc0d716fafb5c6bb8538861dbfbbfa6f711a8449a065de13cee0680b0cb94fd4",
   "labels/objects.json": "e88ec759b4fddf073f003229d88c4b2ecd0ef371e328e39f703038981d9a188c",
   "labels/visibility_states.json": "8eec425f363be7989f97225c4d7b5a50c5a1cec56aa2a39848b9a0efdf7e3459",
-  "source/cameras.json": "5227bf9065e69ed8e6e9c7e98aaf93db584bd11605038f58a00d140bb22ab5df",
+  "source/cameras.json": "0ac6bbcd71ed34f99dc6ef665ef0101f75b0bc26fa15ec1dfedda0a2308e9a07",
   "source/caption.txt": "c92af8ac8e85525e8271ac4f6d4095a3b2657944335a92652b8a9d6cbc8f376e",
-  "source/meta.json": "de0d542cf5406dedc05fd6d2da09311fecc67fe5de5bdf6519e92ccf0de1b499",
-  "source/points_target.npz": "89468147e7e1626e62b1cc819ece74315282d792454896999b135b25a5ee4767",
-  "source/visibility.json": "eb920b9405f643d2b4e4ab8ccdaf0236ee45ebc5162b4f559d7158b4b147afb4",
+  "source/meta.json": "df01f6ed1ed09073520e8ba37de7127bd2217a0aab3e3d8dbd775dc6adf8b726",
+  "source/points_target.npz": "560c659abdd5e459f58e6f9dfc44fe863a912a30d59a0c00ed7458269a290d54",
+  "source/visibility.json": "42fd243adda2030d9236ff139b312979c9466aab00a191188b8602b3903ed0d6",
   "tracks3d.npz": "239326f302ec46bca766226e6627aa68a8afac452587171f55ec2e2aac83903f",
   "videos/view0.mp4": "146b8df625a572c8da11362adaedaa96b0ede79bb70d7b073d1aad69fd667a1a",
   "videos/view1.mp4": "36ae57b14b5956e32d38227d11e2e5d2e4643a78c9eff7ad78589a48a433f466",
diff --git a/Simul-Robot/c001_episode_008070/source/cameras.json b/Simul-Robot/c001_episode_008070/source/cameras.json
index e55c49c58d7edab72c6ae2f770f6d16db3d740fe..bfd9545014445c20152161bda46d82615fddd37f 100644
--- a/Simul-Robot/c001_episode_008070/source/cameras.json
+++ b/Simul-Robot/c001_episode_008070/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_008070/source/meta.json b/Simul-Robot/c001_episode_008070/source/meta.json
index a84790158ff2198ecea382c034f9b16b930377ac..72d4238d22d9de92eac4fc0f0ee4197040b70876 100644
--- a/Simul-Robot/c001_episode_008070/source/meta.json
+++ b/Simul-Robot/c001_episode_008070/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [7, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.02, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 10f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [7, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.02, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 10 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_008070/source/visibility.json b/Simul-Robot/c001_episode_008070/source/visibility.json
index 84f3576689aeec9adceb94e6277bf5c00b2de973..755e8555d345048aea827dec4e4f8e9fc25d749c 100644
--- a/Simul-Robot/c001_episode_008070/source/visibility.json
+++ b/Simul-Robot/c001_episode_008070/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "green_energy_drink": [
@@ -9171,9 +9171,9 @@
       "target": "green_energy_drink"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [7, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.02, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 10f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [7, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.02, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 10 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -16469,5 +16469,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_008090/sample.json b/Simul-Robot/c001_episode_008090/sample.json
index a6b4d17dae0f274599817646b333a0205c2bf75b..be1ee9876e09c06d8df54a23aaa4a8c425122f63 100644
--- a/Simul-Robot/c001_episode_008090/sample.json
+++ b/Simul-Robot/c001_episode_008090/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "096e8d8595e740b3941096d23c41041ff23443d4254dbd7a1a17c35cd248d32b",
   "labels/objects.json": "d53e02c50898dc42441602b277a05ccdd71d74f77cc6b4682e184cf88781abcf",
   "labels/visibility_states.json": "334b34b9df0b2e823817a72323a5d265c0d5b4af2ae8e704512ad3fd756b1890",
-  "source/cameras.json": "a8a0c4700311551c54bc81ab7b88af5f151839f0baf89e1ac4003563361eb260",
+  "source/cameras.json": "8746c278feb413b936edf7d5e275bdcec7602a96d1b3e3edaca3378a073a41d0",
   "source/caption.txt": "870ae3144b1b52174fe58e0139054b901c68da1b8371b229270dc58db0944ad7",
-  "source/meta.json": "ab3977d19e978d6b472393ab332c814685b220b2ddee70e12b69abe914c8a9c8",
-  "source/points_target.npz": "075c92b8725f89b0a2c5689b64fd6511137af5b10e1c9f12930e4a307214abb7",
-  "source/visibility.json": "207767acbfb1212b4365cdaedcdc9b4e5324fe72ef61b2bb9370da9254d583c8",
+  "source/meta.json": "128bc8bf30fcd23b489cb268913837854fcad62a0cb2dc48239dc3b329b76594",
+  "source/points_target.npz": "9dc2e12138b369fdd6e9292968b413cf0000bc8a588f5ceaed79e682d9aa4ce5",
+  "source/visibility.json": "c4792d9e52ab186e633002293a4f9cf63d519883caecf2024184a8d85917ff76",
   "tracks3d.npz": "9489df72e08940c945b870353947415f344ed8ab1ef93d756b4ae717a03d41aa",
   "videos/view0.mp4": "4300886624c9b1ef152359703cd1b77ca7fcdeb9f14e456c286723a95e84889a",
   "videos/view1.mp4": "b6287049cabc9003f87e3317a87e8d30016d6df9764345195f1fdaa04a9d97a2",
diff --git a/Simul-Robot/c001_episode_008090/source/cameras.json b/Simul-Robot/c001_episode_008090/source/cameras.json
index cbd2d23c6fc69abc28430e0da8d7cfb900b67e1c..b1599128545dcee65c5871edd4cf28be991b932a 100644
--- a/Simul-Robot/c001_episode_008090/source/cameras.json
+++ b/Simul-Robot/c001_episode_008090/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_008090/source/meta.json b/Simul-Robot/c001_episode_008090/source/meta.json
index 3040f62d47227c22b42ff9973158925e7e7c715c..fd91821719a0cbcef324963c98d5a352e73179a6 100644
--- a/Simul-Robot/c001_episode_008090/source/meta.json
+++ b/Simul-Robot/c001_episode_008090/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [19, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [19, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_008090/source/visibility.json b/Simul-Robot/c001_episode_008090/source/visibility.json
index fce0762fc3cfbe62103cbea3ef72f5ec326620a1..8c37192b149a750d67d45c14b1ee60951adee60f 100644
--- a/Simul-Robot/c001_episode_008090/source/visibility.json
+++ b/Simul-Robot/c001_episode_008090/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_energy_can": [
@@ -6171,8 +6171,8 @@
       "target": "small_red_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [19, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [19, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -11068,5 +11068,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_008120/sample.json b/Simul-Robot/c001_episode_008120/sample.json
index 751efa0437a2d83be8d8feb1dbfb7773bb6d2580..ad8f23c257ecc68aebb41e80f39c9166c2a7044b 100644
--- a/Simul-Robot/c001_episode_008120/sample.json
+++ b/Simul-Robot/c001_episode_008120/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "937a531527c9aead08adf358139822b3238670c12b95195aff1013cba3c3f482",
   "labels/objects.json": "e30dc202ca76895e8805455713bd9ee7da2ca42c885faaf5109cac8cd94076a7",
   "labels/visibility_states.json": "8858ebe87f568c42b309f262edb664b3d5421a6458d411757ad5b9540b6fc9f4",
-  "source/cameras.json": "c04938ab4df74e9b18a9695f58c6fd4140afe48d4f9893a1a5d7f6ae6776db91",
+  "source/cameras.json": "8d4441790bae2eeeceed8adcc6d0e473a725aec119409861ebd118db46e6ccd7",
   "source/caption.txt": "ca88bbf537565501092aa786014c5922f370cf58f2039c41b397807ae60ec2de",
-  "source/meta.json": "227eb1b1c7b26a95ff94a25e43f669fa80e2d1de2e0d45a0425136e8bba1273d",
-  "source/points_target.npz": "5ad753dd7b121d31341c4f827a55b44098429b36d5fd33ff3fd1bd52e1ccebeb",
-  "source/visibility.json": "633428b91b7a190e0e93bb9be22ff75a14f7b5ea99e3e3e1ce0783c9ef588a15",
+  "source/meta.json": "1805af5e09b839e519f9541b33f400159925faa9ac3323c86125d64bbd2eb741",
+  "source/points_target.npz": "eefbd965f27582cf299c34cded595b4c70e2c944548eda6800383a8f267cc3a1",
+  "source/visibility.json": "c375ddf1190d707b74b426d6f9de06160a0876c5448e174c0244f44055bbe479",
   "tracks3d.npz": "4ed9b570a34a51907705541ab5beeb923e3415b452d21393cb09fe41c3ee9fb0",
   "videos/view0.mp4": "2927515c0e955f933bd6ee851816c43c006ef40c7a37c20c47a938cebfdb0ee9",
   "videos/view1.mp4": "e81a18059ca9c4d1f6da87afd647f0b536b0abc4d48f5ccbc010f0b8c42126ba",
diff --git a/Simul-Robot/c001_episode_008120/source/cameras.json b/Simul-Robot/c001_episode_008120/source/cameras.json
index a4f16cfb8f9d1d5e3b6d187d3347bedef8bae36b..07221e634dac126f8f52f90cbe675030e61a8489 100644
--- a/Simul-Robot/c001_episode_008120/source/cameras.json
+++ b/Simul-Robot/c001_episode_008120/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_008120/source/meta.json b/Simul-Robot/c001_episode_008120/source/meta.json
index ef37a5682d8e29ea18ee4c35ab50585621718c94..0c7f1a75e4d5ace6e90b97264e213cd7961912fd 100644
--- a/Simul-Robot/c001_episode_008120/source/meta.json
+++ b/Simul-Robot/c001_episode_008120/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_008120/source/visibility.json b/Simul-Robot/c001_episode_008120/source/visibility.json
index 0a09378a0612fd9d9aaff6f9309217e1f07330e3..07a08079fafb80e4ce4478d8afdcc23d6c0d0e2c 100644
--- a/Simul-Robot/c001_episode_008120/source/visibility.json
+++ b/Simul-Robot/c001_episode_008120/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_seasoning_shaker": [
@@ -5118,9 +5118,9 @@
       "target": "peas_food_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9176,5 +9176,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_008210/sample.json b/Simul-Robot/c001_episode_008210/sample.json
index 2128a84a8d253d611c2f82419a1b7f8d99b4e284..9a72da7a5f02b70f91d42d55b77b0ca706a208c4 100644
--- a/Simul-Robot/c001_episode_008210/sample.json
+++ b/Simul-Robot/c001_episode_008210/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "608da44cf79e3a95efaacf31566224b9a048eafbbbbb603f355457c1ceca2a99",
   "labels/objects.json": "b8a3d17db93730606ef7a3193e76ae44f5dcfb9f9031c504f48e62fd9cdcdc94",
   "labels/visibility_states.json": "81eaf6805230ef914aae3cb3229585591cf52b146bfc816dd87c869cdd344dc3",
-  "source/cameras.json": "c99e6c272cdf1808b19337af32c5f49e5de058906e3e913ef353a722c281ba21",
+  "source/cameras.json": "580c0d7bde5670865b233b4c46f9c3f6e867b8b003aa5aafa1e9ff16597592e2",
   "source/caption.txt": "5b04dbd3e980330b3f9fc3336bc8f39cbef97b51da449b1ca25c1d93b62968ff",
-  "source/meta.json": "a13990750ff7fb8658b787b2d95efa5b7b8919e6768ecf4a14d27a9c4f8204f2",
-  "source/points_target.npz": "fd67b3c4c3f49106f5d4f3ed32f2b90b0547e9d71d12e94d96505fc5a7232f47",
-  "source/visibility.json": "9bfbb6c448a48863b9b39bfdaf64363bd37cc7269e1d65dc4ed177d1eddc1b2c",
+  "source/meta.json": "9c37318e1d5efa04d53e9a11e0e2aebaa55126a249b87bb1a80a1f7d0752be22",
+  "source/points_target.npz": "2bb6361d6cd866880d0ff64344ca5a06eb66e55e7d1f1bcff3d0a8b199fdcc6b",
+  "source/visibility.json": "43cec1ec798b57e29217d760c85f61d7ecb80112c8ec7315f738e34d2c533cb8",
   "tracks3d.npz": "72379e65e0defc1215624df38c14bb50f21a5f898bbdf74126dac7fb24632383",
   "videos/view0.mp4": "d6d8d09c19322ae0aca932853fc81e95ae08992ddc93ea3fb2e5f13fe62111b3",
   "videos/view1.mp4": "ce102b82a326b60aea5c7af834a22177dba6c43c10aab498a30db606269a2ff0",
diff --git a/Simul-Robot/c001_episode_008210/source/cameras.json b/Simul-Robot/c001_episode_008210/source/cameras.json
index 19529b8e7244e9c270bf9c667bb1b66f5bc5ddd4..16c687dd7bb55e26875e43ac247dc980429492d5 100644
--- a/Simul-Robot/c001_episode_008210/source/cameras.json
+++ b/Simul-Robot/c001_episode_008210/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_008210/source/meta.json b/Simul-Robot/c001_episode_008210/source/meta.json
index 8b0437e036a740610a4085b2436a15898d34464e..b4aed071f19297640080a7f75f18d0760f924a54 100644
--- a/Simul-Robot/c001_episode_008210/source/meta.json
+++ b/Simul-Robot/c001_episode_008210/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [7, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 11f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [7, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 11 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_008210/source/visibility.json b/Simul-Robot/c001_episode_008210/source/visibility.json
index 181d8460139f0244ce31e128946594bfa3ad8599..75ea75ac5584ceb6fe11342bf69b86222dcf44e9 100644
--- a/Simul-Robot/c001_episode_008210/source/visibility.json
+++ b/Simul-Robot/c001_episode_008210/source/visibility.json
@@ -46,7 +46,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_seasoning_shaker": [
@@ -5121,9 +5121,9 @@
       "target": "small_blue_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [7, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 11f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [7, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 11 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9179,5 +9179,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_008240/sample.json b/Simul-Robot/c001_episode_008240/sample.json
index af3daecccf22fc3808dd4862628daa07411a00dd..67673731f7b868a571156d943cef623bc982a5d9 100644
--- a/Simul-Robot/c001_episode_008240/sample.json
+++ b/Simul-Robot/c001_episode_008240/sample.json
@@ -220,11 +220,11 @@
   "depth/depth.zip": "224bd26d7fc043aaf84ce8c3be67cb16d821edcef208597c44d3f08a977ee950",
   "labels/objects.json": "7f243476fb6fe30d68d3105cae57d385a18f0fc226e769808643a1d08a07ca4d",
   "labels/visibility_states.json": "ae8bab482a9caf59a16f387ed672cab439f381b6059755689b36388d700e79c9",
-  "source/cameras.json": "fbdd69e2b2dbfa8e57d8b544630a744ee700bd21eb846d5e55d212f855a27319",
+  "source/cameras.json": "67ad915449d342bdfc090b7541150b6f70e6b4b18567a7aa7e5e330fbb4f8f85",
   "source/caption.txt": "c8c1c6a875c0b30bc347c3ae1f128ec809ab6590ba2a790fcbbb1e26bb1f37c0",
-  "source/meta.json": "b693abf45c7042efd370315934bdbd576860255cdbdf12a68cc9176ce26d0d2e",
-  "source/points_target.npz": "83d21d97f6d52187889c544a8669c52b2d87d051ad939cde9b6aff549194456f",
-  "source/visibility.json": "4637a5daaec3558209625da8d139d834aa3ed64b26d45af7349ed14fbf0b3014",
+  "source/meta.json": "288b2bb912f49e357b97585ea64df397cd9dd9c77d0f1669701432a2516a0f46",
+  "source/points_target.npz": "5b664f29ee0c06c8b1cb2fc41a2248a8ccf8f9df44b6c8d6543dbc309beaa53a",
+  "source/visibility.json": "baa71c951017c7b1405a09e7b03bd441fe3dcb50423e9ab4f7a3b4f27db06b75",
   "tracks3d.npz": "664dcaba7d8a0c4c011325bc1141fa4cfcc467bcca632ffae0c7dec5f62ff343",
   "videos/view0.mp4": "667191d92647a34ab18bdd0413ec7fa25a455cc66539096f83fa55b6fab3efeb",
   "videos/view1.mp4": "76ac38af188b95cd2788a132a89a791729431bf024bd25375ad02b308fbc3441",
diff --git a/Simul-Robot/c001_episode_008240/source/cameras.json b/Simul-Robot/c001_episode_008240/source/cameras.json
index d9778224767e0aba78d1e0253d723262d90b0491..aad4a82bf4f30bb75fb69ec01cc3cb66401ddd08 100644
--- a/Simul-Robot/c001_episode_008240/source/cameras.json
+++ b/Simul-Robot/c001_episode_008240/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_008240/source/meta.json b/Simul-Robot/c001_episode_008240/source/meta.json
index b2366f8bcd1b9313160c71cfcae817e6692c9199..766e82bf9ffae7dfb8bfd61724d813a388971b7a 100644
--- a/Simul-Robot/c001_episode_008240/source/meta.json
+++ b/Simul-Robot/c001_episode_008240/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [2, 2, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.0, 0.0, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 4f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [2, 2, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.0, 0.0, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 4 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_008240/source/visibility.json b/Simul-Robot/c001_episode_008240/source/visibility.json
index 989184432f716f162b2979e1f595d675ef8fcbf5..a92747f7f671f732e1c0d82060bb4e08982f4578 100644
--- a/Simul-Robot/c001_episode_008240/source/visibility.json
+++ b/Simul-Robot/c001_episode_008240/source/visibility.json
@@ -41,7 +41,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_frog_can": [
@@ -4955,9 +4955,9 @@
       "target": "blue_frog_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [2, 2, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.0, 0.0, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 4f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [2, 2, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [0.0, 0.0, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 4 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -8639,5 +8639,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_009340/sample.json b/Simul-Robot/c001_episode_009340/sample.json
index 52548d6ecf67bb09c414fe89faf3fbb081dcd561..6f25f1ad7b738745e37960269b8e5fdcb2692af0 100644
--- a/Simul-Robot/c001_episode_009340/sample.json
+++ b/Simul-Robot/c001_episode_009340/sample.json
@@ -220,11 +220,11 @@
   "depth/depth.zip": "01d4fde919bc5b45bfa3eeaca964ab212f6a5cf8afad6cc45f3786a49a126ec8",
   "labels/objects.json": "f3c8ccd2473b8f1904bc519ffce0a486e34f3b2ea9dfe8e7faa12fcd949f5e07",
   "labels/visibility_states.json": "349c7d06a0734c1df5a542bfd778cad280837abe00e339865e659ad576931f17",
-  "source/cameras.json": "223aed0cbb1073a6f90d701e6d77bfbe2bbfbdf514d35fe1d46b296860a87360",
+  "source/cameras.json": "17a8e1a7d2da10c9b0727677fa923aff46e3192971721c01f42037c55ae395e0",
   "source/caption.txt": "27063a92d6b9d9063444f7767c1a36291c3023cd30b103dc274f3fb2c3fe5391",
-  "source/meta.json": "21f15c6935307733b829f691a4dad9d5b906325a943b9a765b88ed77a2d7e565",
-  "source/points_target.npz": "e62e1c66adead3b14ed49c7fd257e563b5c15ac54b19396b100bfc5f654accae",
-  "source/visibility.json": "3e721d25b62ef81451973f3ab53737dd6da73859711e93e902e38f920b6daaaf",
+  "source/meta.json": "2b0ea4b4ec5a0d7c83dbb88254f45319bad22904e2aa23e79c355c58e3faf0bf",
+  "source/points_target.npz": "e1600a7c53dec5d1d20a0aae1bbab637899d2709334774e42d6ffddd8795e988",
+  "source/visibility.json": "ce19a085beff37cdba6ffe81e8372ccbd54df733bf0c4a415d4926c24449f12e",
   "tracks3d.npz": "604fd301dd6ba8fd7fcf8f66ae67545f68533191493e75d32fabf6ab403289dd",
   "videos/view0.mp4": "526d260b0325d66a38395e9e492f47ac99ccacd5b550de1e4a789d697d433c50",
   "videos/view1.mp4": "e05f4a265908890649acebf45b39ea4bef5842fec67c1072f3ccd484e1227752",
diff --git a/Simul-Robot/c001_episode_009340/source/cameras.json b/Simul-Robot/c001_episode_009340/source/cameras.json
index b5819f952d9b3efab3b49667a535b79c14ffd5a9..5658a2ce4e689dab454e5f59fbb143d45d0446ee 100644
--- a/Simul-Robot/c001_episode_009340/source/cameras.json
+++ b/Simul-Robot/c001_episode_009340/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_009340/source/meta.json b/Simul-Robot/c001_episode_009340/source/meta.json
index 0deb08ba6e394e1b3a830f0ec12dd93faff23fbd..532c2a69556d30dc3394e21730e8449494248eb2 100644
--- a/Simul-Robot/c001_episode_009340/source/meta.json
+++ b/Simul-Robot/c001_episode_009340/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_009340/source/visibility.json b/Simul-Robot/c001_episode_009340/source/visibility.json
index 025492c97daac67302181cee67a1fe1ecc8f084b..f332c35fbc90b3c55421e1d2eace2f24e39b8407 100644
--- a/Simul-Robot/c001_episode_009340/source/visibility.json
+++ b/Simul-Robot/c001_episode_009340/source/visibility.json
@@ -31,7 +31,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "kraft_coffee_bag": [
@@ -1711,9 +1711,9 @@
       "target": "kraft_coffee_bag"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -2567,5 +2567,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_009370/sample.json b/Simul-Robot/c001_episode_009370/sample.json
index 0ec8192b9654be99ca753d9e673303177a1edc10..2aee5ec7719aa40f4c78911ed7b64fade5f85023 100644
--- a/Simul-Robot/c001_episode_009370/sample.json
+++ b/Simul-Robot/c001_episode_009370/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "4d70f75642875a9010f7a15e71f730177a8960235c6909a194aaeb4a859aa8ce",
   "labels/objects.json": "d1d5ca2eac33bec6ecf10b6317bf1806cea5458a56c0e1d30c5d1649b616c14b",
   "labels/visibility_states.json": "5c187ce4d6e63e6eb75a1678107fe4b41ef47eb2ed9346e78323a773ddec3124",
-  "source/cameras.json": "756173aed34f5adbaf500b2cf90f7ee9574d9a99fe822042440bf694470be306",
+  "source/cameras.json": "e94ae22c126d9b01b7182f15e20bdcbd3448ff16ae734829702e9a0c4ca3d55b",
   "source/caption.txt": "546ad37157dd6443199b862ff9f78b47b728e8022b50f19648df3aab0f1abd57",
-  "source/meta.json": "e4285175473c8ea3254e0ee4a60da8551f1a6953b2e49424e626a1d772059820",
-  "source/points_target.npz": "b59085d5eb6e4acebf32f4872793ae825ee78f8edcd2b93f779a55dc3e2b413c",
-  "source/visibility.json": "4246520af15b698c615859c983f8abe678f645dfbb4e9b0f9bb996214a228541",
+  "source/meta.json": "789f3b11a418a9f91104a484c37e5f40a5ee08320409f00c43f77f40bc8b9207",
+  "source/points_target.npz": "8530f4dcc9727fc62900eb96b961722a12a489caf9e2fc790e0b0c3baf4fde31",
+  "source/visibility.json": "fb42a07ccf4f3bd1f90f9c3dc9da51f610e33bf305f4291153a497984ea9e592",
   "tracks3d.npz": "37847a774126a3a35a3da15a8f4c7bec527af00ccbb0c0f73fba7edbcc92d4af",
   "videos/view0.mp4": "6f76560a9ee9b9527eea3e34f58dfa23956ce08d8dfde716cc2c79327e494077",
   "videos/view1.mp4": "17bf3f192b5d3d1a3a935112edb96ff5a129e4cd5233ac985c1a929120bffa14",
diff --git a/Simul-Robot/c001_episode_009370/source/cameras.json b/Simul-Robot/c001_episode_009370/source/cameras.json
index 4236bbfad3259cd0370180f34c7130c630d22437..eea7df123f01e7503bc913290f1028fb9be87b2e 100644
--- a/Simul-Robot/c001_episode_009370/source/cameras.json
+++ b/Simul-Robot/c001_episode_009370/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_009370/source/meta.json b/Simul-Robot/c001_episode_009370/source/meta.json
index 91084c80b9357ac452461cfd4c01f7d243a81823..de4a11c7d0218471616d5aed74ebe2123ababaad 100644
--- a/Simul-Robot/c001_episode_009370/source/meta.json
+++ b/Simul-Robot/c001_episode_009370/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_009370/source/visibility.json b/Simul-Robot/c001_episode_009370/source/visibility.json
index 6b8cc5c131007e33b8ffa02ee2042f1319ebb961..ee32e43ad4f3d9220e2b3b069e8ca93c0d28fe5b 100644
--- a/Simul-Robot/c001_episode_009370/source/visibility.json
+++ b/Simul-Robot/c001_episode_009370/source/visibility.json
@@ -31,7 +31,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "blue_patterned_mug": [
@@ -1711,9 +1711,9 @@
       "target": "blue_patterned_mug"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -2567,5 +2567,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/c001_episode_009940/sample.json b/Simul-Robot/c001_episode_009940/sample.json
index 3114749e237e4729d1352ac170615490b2838aa2..07a194abe5ffc659c8fb41b9ce3db0ede1df1f66 100644
--- a/Simul-Robot/c001_episode_009940/sample.json
+++ b/Simul-Robot/c001_episode_009940/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "e0f149e497f93eb9fa718276964402fe809f3960d71a9c8a47645883b48bd306",
   "labels/objects.json": "3cad7a68d1a9fef0cf17e2883e270e91d34234bf111d6d3aef428ac0de525324",
   "labels/visibility_states.json": "30b4446d92b18c1b917ed9c1e8fc43bebf40b8b0a99b3c8425b6b5c763fe49fa",
-  "source/cameras.json": "6ee7df67c90abad0b4b7db939b1d02134db90302ec2b105fb6a8b1755c0e29b9",
+  "source/cameras.json": "b9cd915efb00de552b84b8ddbf27860b8b133888ce11eb342bcf9785621508c4",
   "source/caption.txt": "09289b7c5705463d9d808125ddda789ed19db7304c5222f989b6ef694a1ac5ef",
-  "source/meta.json": "3ed290b9ef4522c94af7a55591c47972c4f93460dfb4049e763f5e449061ecb3",
-  "source/points_target.npz": "0d717fd6bbe052adb5c742bfc66a8bbaf666f41b57eec29d438d8299c1714c7b",
-  "source/visibility.json": "5767ae4cffc8f5a8b40711f2fddfa826bce3924935123a3d78d8b0fd29cb9fd9",
+  "source/meta.json": "be4f564d263628218e64fd7239d6190c54c54b0338978847568409d813794b22",
+  "source/points_target.npz": "6ab3c8eecdbc7595d0eb18d748d12413a12273acd4b5a7fe02bfcea119cd9c3f",
+  "source/visibility.json": "66d1cc66d83afca8ce9c670cfadb3d8a5f8ff6d7cabecef5d18e2d0dade9e165",
   "tracks3d.npz": "b0685886d4af5dcce9aee7c5599b96e2a04f47af73bf71b81d372c0d299f6f36",
   "videos/view0.mp4": "8347fcdd598757ed05a86309f6706e53974948d6214a1b9ec3855a9555d1e63d",
   "videos/view1.mp4": "5b84997bc0a15938900c0e64f11898b863180555f4464de80f7e1b9373c1bc0c",
diff --git a/Simul-Robot/c001_episode_009940/source/cameras.json b/Simul-Robot/c001_episode_009940/source/cameras.json
index f658f9e4180754d3273d84d41ac0a66656be6750..087b1c0f65bbd5e5593c398045f3dd797557da86 100644
--- a/Simul-Robot/c001_episode_009940/source/cameras.json
+++ b/Simul-Robot/c001_episode_009940/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/c001_episode_009940/source/meta.json b/Simul-Robot/c001_episode_009940/source/meta.json
index 205c77b74ea5d05adc621f3d84acfb1bcf56aa82..99e227ba99d182dbe6bf899cebb41e451c16207e 100644
--- a/Simul-Robot/c001_episode_009940/source/meta.json
+++ b/Simul-Robot/c001_episode_009940/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_009940/source/visibility.json b/Simul-Robot/c001_episode_009940/source/visibility.json
index 1996eb8e49c422185454b453d786b448a25c8144..2754eea662b5bb48e82a4bc12e87397809902f51 100644
--- a/Simul-Robot/c001_episode_009940/source/visibility.json
+++ b/Simul-Robot/c001_episode_009940/source/visibility.json
@@ -31,7 +31,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "kraft_coffee_bag": [
@@ -2091,9 +2091,9 @@
       "target": "kraft_coffee_bag"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -3137,5 +3137,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k001_episode_000100/sample.json b/Simul-Robot/k001_episode_000100/sample.json
index 3b59e7d9ee0b39bb5dd30a4a92f7ae6aa2767328..5da6eca4e9ecc683565bf2853f1f98cdb98b9470 100644
--- a/Simul-Robot/k001_episode_000100/sample.json
+++ b/Simul-Robot/k001_episode_000100/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "7b36ac6c7e1e126a532faeb7486cf3924c93719977e469463451208431844258",
   "labels/objects.json": "ec818f04892a8925a64709c82d0ec0870b080e36902913e490facb2f822c304f",
   "labels/visibility_states.json": "f635da5286a9db2cdf6ceed81d16d6227841ea967486a4665164d4d1c4811b0c",
-  "source/cameras.json": "38e85cc74cb2b384b2dba1f084b3dad4b652a9d24f9e1c2e5254a69c48966fff",
+  "source/cameras.json": "de413afa2d1c44096c2dfd8ace28778636b480d6d3dbf6f744a240a7605df50d",
   "source/caption.txt": "22619b1faf09e073712306571e63fed2ebace4a581e2fca547f726eba3b16d8d",
-  "source/meta.json": "f219954afd5d8545226738f846ed7b95f8b3a88b850b3a341a79d3dfc742dbaa",
-  "source/points_target.npz": "432704df071c2a56692c2e86b0604eaeb53bf6a16fe54e38afdb16ff71d4dd45",
-  "source/visibility.json": "6aed880c924348df2bd5dd8c33e660ff160539a2022f55a2c5096b2cabdd4326",
+  "source/meta.json": "1e4075fa3fdef8e97808a69ca731c2f8c8a9cc19d154da71909034c3dffbc546",
+  "source/points_target.npz": "8d869a4453353bf58f965bca558394a0b8b443c5af7fb35e2ebc14871f449b9b",
+  "source/visibility.json": "76e80a92d4959045bef1c3ce7cd68715aad752f5247223375d6bdc50bf134aba",
   "tracks3d.npz": "ded801f7e58c7f1f864b2fc260f218e5df5ab6ff0ac0fdf4c277258eb29333f5",
   "videos/view0.mp4": "7fd0de85ef3c997fbd0406aa6dd3d1a3058382204952ffa529bb467babec3a0d",
   "videos/view1.mp4": "6af098fc47498b301a61b8fa90f6f877ab3dbc741dc28f9fa22e5771b7141622",
diff --git a/Simul-Robot/k001_episode_000100/source/cameras.json b/Simul-Robot/k001_episode_000100/source/cameras.json
index c456ad5bfcd6a11b51d1d5eab89f9d3cc5321670..825af0e73c06b6368418f7a278520b1cf1570954 100644
--- a/Simul-Robot/k001_episode_000100/source/cameras.json
+++ b/Simul-Robot/k001_episode_000100/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k001_episode_000100/source/meta.json b/Simul-Robot/k001_episode_000100/source/meta.json
index 2507dbafd587d22b5e5c2ea2736a547f53dc2a2f..9f0bd3da61c29eca35d79838887df3f330b09791 100644
--- a/Simul-Robot/k001_episode_000100/source/meta.json
+++ b/Simul-Robot/k001_episode_000100/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 13, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 13f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 13, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 13 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/k001_episode_000100/source/visibility.json b/Simul-Robot/k001_episode_000100/source/visibility.json
index f5d2e4f8a755ccb166e0ed3e6fce70cbd37187b9..47db604fabce5e5e1605f12eb253cb857c512283 100644
--- a/Simul-Robot/k001_episode_000100/source/visibility.json
+++ b/Simul-Robot/k001_episode_000100/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "small_green_cube": [
@@ -8293,9 +8293,9 @@
       "target": "small_green_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 13, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 13f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 13, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 13 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -13803,5 +13803,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k001_episode_000550/sample.json b/Simul-Robot/k001_episode_000550/sample.json
index ca140b1586efd53a657998d08cb2b04cb7ee5ec3..9756bec28ff6e5c38ff2af32a61d6a6716536e10 100644
--- a/Simul-Robot/k001_episode_000550/sample.json
+++ b/Simul-Robot/k001_episode_000550/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "7cec6406ba4f97635ca4928ed6ee4bf3aba0bc15f60dd42dbeb022a47708a696",
   "labels/objects.json": "80fe5a0659f421469ebf354ac0b8518f87260f92de6b09ca30d29defc00c98bc",
   "labels/visibility_states.json": "fef22b4e4fed8c7284fe9e63de065123b20df3d22af308715901516e93cb0722",
-  "source/cameras.json": "2e982f5a2830e8d9f329c1571b4235183bb92c17d0e4cb758bc5d0d657472701",
+  "source/cameras.json": "4fc1f413725218d00ffe966b3f370170d386922977741b08b2b662484b27db12",
   "source/caption.txt": "3d00e11a3fe72c1ffa44c1bdeaa64569e453c15bdc8b0caa970a47392aec4f9a",
-  "source/meta.json": "f0b97592277f4dc4eb2888fd80340b6be19788afdb1e2f0e8593135d0f49383e",
-  "source/points_target.npz": "cbea3b25738524a853aa998ce64917844816e807d8d3b0d428c912c306769478",
-  "source/visibility.json": "e47e329b3da22314e06823a55d7577c1b93c03b868d9d9743135a725b055dc62",
+  "source/meta.json": "4dafc431c35b7f7fd820bb4d2b9745385a41f4ebcc95d7823f400204bed52c3d",
+  "source/points_target.npz": "36ae1ab80b8db2e1725e7fd1ae706fff70f94df07da959d68340498a620b4188",
+  "source/visibility.json": "75d6a3f9769ad0af73df4191dfd1437b3c2faf3833b18606718bd8359411769e",
   "tracks3d.npz": "a1856652a68e2f4a8d2dc57bad64e8d9538626baa974e0ccb952fc827f299ff6",
   "videos/view0.mp4": "ce02040a9ec5dc1739ae397f514d13ab3cbcc4f858f95cee439ce24a7793e355",
   "videos/view1.mp4": "dd38278f332b33eeb56b4738a00464c1dedee3be898afab72a2e25b1c77425b9",
diff --git a/Simul-Robot/k001_episode_000550/source/cameras.json b/Simul-Robot/k001_episode_000550/source/cameras.json
index aa08d6532f4f7a2d46caedcdb58bf3643c0eee35..6b11aa37fe785572e2538db5f2e38b171cf55155 100644
--- a/Simul-Robot/k001_episode_000550/source/cameras.json
+++ b/Simul-Robot/k001_episode_000550/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k001_episode_000550/source/meta.json b/Simul-Robot/k001_episode_000550/source/meta.json
index 92cda473f63cea73e8ebbba1e50bdc240b2f1bb5..09a7325e203e689497762addeb366fac772477ca 100644
--- a/Simul-Robot/k001_episode_000550/source/meta.json
+++ b/Simul-Robot/k001_episode_000550/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 2,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [2, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01'] 의 비가시 구간 타겟 변위 [0.0, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [2, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01']: [0.0, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/k001_episode_000550/source/visibility.json b/Simul-Robot/k001_episode_000550/source/visibility.json
index 737c2472db876ee1d9ca8c19c40e28d1d63b55ca..4eda38440db4944b73937f01a49a116291f25d6c 100644
--- a/Simul-Robot/k001_episode_000550/source/visibility.json
+++ b/Simul-Robot/k001_episode_000550/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "small_red_cube": [
@@ -5296,8 +5296,8 @@
       "target": "small_red_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [2, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01'] 의 비가시 구간 타겟 변위 [0.0, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [2, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01']: [0.0, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -8805,5 +8805,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k001_episode_000570/sample.json b/Simul-Robot/k001_episode_000570/sample.json
index 8b6eb35ee171efb0dce946a08da6e38bcde908e1..3d9a9e109ceec2222a6140257118dde4ddd89493 100644
--- a/Simul-Robot/k001_episode_000570/sample.json
+++ b/Simul-Robot/k001_episode_000570/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "da4b6f34c45a62e341f3a86b280c82fd870c8e0810b989608e25c2691c5b1afe",
   "labels/objects.json": "99f8edac3546371dbee6c0425fc15529effe4587206d6278535a238c000668a7",
   "labels/visibility_states.json": "a2eb6b6b9b71d6163416b90615555c48a3e92ab5f051c5d343cc6cd1002e20db",
-  "source/cameras.json": "5add1af3af2039d1139594df79f91b4ab23dfed883af5464ce4efab96c097bce",
+  "source/cameras.json": "d6a2354a705cbb4a691a9178f4f678039ce598ade38c62623eb666ab0b3a8db0",
   "source/caption.txt": "d03f074556e0b60b28a9ccbe325ee0b7804819c1d7359e52a6413246282baf68",
-  "source/meta.json": "31aa663be392585f7e960078e3f0b9b896418c02abcfeadd52f2c664266e8d83",
-  "source/points_target.npz": "93a3ce7ac990218d99217aa15e893533d120696efab4aab5456d6671b9567366",
-  "source/visibility.json": "349a778eb96b047497ef2a03521b147f938f16f96aea42159ae3260063b38318",
+  "source/meta.json": "01eb1ff14a5fb9899402f2035ceb86afc12d8c5182f4cbc1962e9759e47b2475",
+  "source/points_target.npz": "bddd2faaf4466cf81f5052ef458c8d0361221d3790b95a32dd8b4d091182f9ee",
+  "source/visibility.json": "dd9594bbcc4b2a3c1056eccecdb1587b8bf12baac4220904c6b8b484319af40b",
   "tracks3d.npz": "8d847c32dedbdede745f8d3c78b16e30c74542f072aeb1fa093e8baf35a5394d",
   "videos/view0.mp4": "1f79b9b34e98318100a7bfcedabd6f3c4566e7d046c0bfb3c826b390f9453de9",
   "videos/view1.mp4": "5f754b9a072a6af98564334c672a20a1a808dcbac324b07329486f25e4774fb4",
diff --git a/Simul-Robot/k001_episode_000570/source/cameras.json b/Simul-Robot/k001_episode_000570/source/cameras.json
index aec04c0bc000c73fbceb5970be80d241a6f2242d..a7d7f973897eef3d0a98750057b7f5f9335ecfcd 100644
--- a/Simul-Robot/k001_episode_000570/source/cameras.json
+++ b/Simul-Robot/k001_episode_000570/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k001_episode_000570/source/meta.json b/Simul-Robot/k001_episode_000570/source/meta.json
index e342b8ec1cc3295d18ec8d9feaccec006e8b0629..9471256012b310ebd638d595ea96b3d311573b19 100644
--- a/Simul-Robot/k001_episode_000570/source/meta.json
+++ b/Simul-Robot/k001_episode_000570/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [14, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [14, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/k001_episode_000570/source/visibility.json b/Simul-Robot/k001_episode_000570/source/visibility.json
index 0e85e4169cad26c53241a6552753df13c3fdddde..73f0ef2f6ee2c716c37c768073dccfc9fa4e5269 100644
--- a/Simul-Robot/k001_episode_000570/source/visibility.json
+++ b/Simul-Robot/k001_episode_000570/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_plastic_tote": [
@@ -5563,8 +5563,8 @@
       "target": "small_green_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [14, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [14, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9252,5 +9252,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k001_episode_000630/sample.json b/Simul-Robot/k001_episode_000630/sample.json
index 89503497110279cb53f2cc709197b60e600fef35..b55724325c1b6f776662b7a48be48d834b479d84 100644
--- a/Simul-Robot/k001_episode_000630/sample.json
+++ b/Simul-Robot/k001_episode_000630/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "d9d7a6ed80e55cddc6ea3fa8d97b564a5461f09fc3fc21f2457aac064acfd76e",
   "labels/objects.json": "ffad6892fc386ee2985db2276d09df94c0c83ead572dcba9558934223aa62b80",
   "labels/visibility_states.json": "0710f7cd8bd488b0a4fb7ebcc5434f6fecb3baad6fd3dec29e2f42a9e2fd6d94",
-  "source/cameras.json": "875d11baf1b89a853e3adc742967763369893d2a9166bd482a6ad900290f91d0",
+  "source/cameras.json": "12964d4b484ef5eae5a8821ba408d310e2ad948e0c7e319e7acd71e164313642",
   "source/caption.txt": "e063abe7a48d9fca523c6353e7c7eb7b400c080a6289a4bdacae8b6910404af5",
-  "source/meta.json": "784278186e0e532c61ffe68a3e4b15aac12dd69e8cb3b15e8ff459e875e1967c",
-  "source/points_target.npz": "29600655aa8e72404931763c9349584d14f4053cf156a9ae3220bb3b40e263d6",
-  "source/visibility.json": "fa27ee4b7d6ce35a4f6c5e6660ac5a2a665f38a4720ee425ffbe7fb3c70224a8",
+  "source/meta.json": "65291ebdbed2070523ddb020142e945055d28a4c78d25646990eb6a1171005ee",
+  "source/points_target.npz": "189a38b7270a8aaab9e5f0b31a8994182a82146642a709d67f41eb707411a16a",
+  "source/visibility.json": "21268d3e641f284cee350bfbacc75ca1b6404a4a95ec4417f4f7de425b2b83f6",
   "tracks3d.npz": "452c331e645777ed734ec808500bb3eb4badf2cbbf010f0793bb9c5cfaacf1f0",
   "videos/view0.mp4": "4e93a5c6bb522b87eff45be5136cf76860ceede6bd7662e28070c41a3f7cc1c1",
   "videos/view1.mp4": "7c89f5a5d5d8eef7a9d177f76a3bfabefef367aeb06990f21a5cd51becb45168",
diff --git a/Simul-Robot/k001_episode_000630/source/cameras.json b/Simul-Robot/k001_episode_000630/source/cameras.json
index 9ea1aaa6d089ba621f4cc0bde7a45052c702cd9c..20afb2fe62ff1354e69919fa930788cf401a9436 100644
--- a/Simul-Robot/k001_episode_000630/source/cameras.json
+++ b/Simul-Robot/k001_episode_000630/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k001_episode_000630/source/meta.json b/Simul-Robot/k001_episode_000630/source/meta.json
index 7c0bafb0542910f81474fa98c7e78a060b9a7bd0..03555726c29464e49c2c4e1473047a4079319a59 100644
--- a/Simul-Robot/k001_episode_000630/source/meta.json
+++ b/Simul-Robot/k001_episode_000630/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [5, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 9f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [5, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 9 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/k001_episode_000630/source/visibility.json b/Simul-Robot/k001_episode_000630/source/visibility.json
index 89df6b9dbcb78fab354f31963941758769ad7e28..8cf1026913408a0bc525767e9cbcc2d4822fd6df 100644
--- a/Simul-Robot/k001_episode_000630/source/visibility.json
+++ b/Simul-Robot/k001_episode_000630/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "small_green_cube": [
@@ -4873,9 +4873,9 @@
       "target": "small_green_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [5, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 9f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [5, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam01', 'cam02']: [None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 9 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -8103,5 +8103,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k002_episode_001460/sample.json b/Simul-Robot/k002_episode_001460/sample.json
index 3a1617d1a63dbd122e00d996bbf774e060dedc31..91a7079fa51d2bcc5367af5ec104b4b274b43655 100644
--- a/Simul-Robot/k002_episode_001460/sample.json
+++ b/Simul-Robot/k002_episode_001460/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "f75517b24032ad84483265580e3d5cb9833a893abfe6a69d8ee26f636ebcc19e",
   "labels/objects.json": "a75cd6713b89fe5be7b66c6d8f0baf4548b4868b76282299eff49deb16bae278",
   "labels/visibility_states.json": "85decd91f3a0aa8ad4e136deade887f4abde23c7df0f63f5ab1095fea50fe31c",
-  "source/cameras.json": "73af0f9e65de9c09be06419583b82db0f8bd75b51407d4f084ecc5dc45fed419",
+  "source/cameras.json": "c29885856eb58ac6839deebf88e7dbcdabf527447ad45195f654bb91e88d4fb4",
   "source/caption.txt": "7e3a429654a3468a155eac7646f960dcd917153d63c4cc16112dbdc6a338fc3a",
-  "source/meta.json": "b75ee65d2eaf536d8a53deb04cf1431bcc89c19f8dcf5248b76ea1638b168415",
-  "source/points_target.npz": "a3c7517f6918809f64af9aad62d483710074028c9e5d4d18a3ecad46fffdf02f",
-  "source/visibility.json": "83179cfc88feb3b2f7f6280884f886cd71fddb202ae0874d71246f69f174b9bd",
+  "source/meta.json": "a70e51a3c235ca2fb95091eb78abc62e3d356a31707288ca6be4064ac6016543",
+  "source/points_target.npz": "4c1e5dbbf28b4ddc4b0cef937fa4486c41c2128e4d0611aa9f70c3b725cba2b9",
+  "source/visibility.json": "2444cdf0cf07a819acded07efd3f45c5bd69b449354be1fcf95b6f5aef7d5fea",
   "tracks3d.npz": "a3d6826ea5222a4e522b35d43984de03131fdcada2d50fbe902f8063f1f262d7",
   "videos/view0.mp4": "cad5fb01019ab08a53a80dc27de5c2306d2a7aead936b801901fccfc9008373a",
   "videos/view1.mp4": "0c8477029ee84313eccf4d8fab7dc91fd0061c01d0d6ce8b8037f6147ffefac1",
diff --git a/Simul-Robot/k002_episode_001460/source/cameras.json b/Simul-Robot/k002_episode_001460/source/cameras.json
index c9d702b7003538039229a61cebd1d58153c66654..30138ae504b3db1de9303ed1ada0cbb68eb35485 100644
--- a/Simul-Robot/k002_episode_001460/source/cameras.json
+++ b/Simul-Robot/k002_episode_001460/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k002_episode_001460/source/meta.json b/Simul-Robot/k002_episode_001460/source/meta.json
index 10fd62078a27b4e1f3cd471f02e6b7f2673a582a..d5dd4b91de2b0edc2f1f0eced0789a38455a42bd 100644
--- a/Simul-Robot/k002_episode_001460/source/meta.json
+++ b/Simul-Robot/k002_episode_001460/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/k002_episode_001460/source/visibility.json b/Simul-Robot/k002_episode_001460/source/visibility.json
index acebcfd017c212a1c628f9670fafa8ebaa113814..528027582cd708b9a29453826ddc971cb45d9f71 100644
--- a/Simul-Robot/k002_episode_001460/source/visibility.json
+++ b/Simul-Robot/k002_episode_001460/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "floral_cat_figurine": [
@@ -6100,9 +6100,9 @@
       "target": "floral_cat_figurine"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -10150,5 +10150,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k002_episode_001560/sample.json b/Simul-Robot/k002_episode_001560/sample.json
index 04532b79f22e2a835414d4b9cefe9504e6fddf8a..2bb1df832fc1d9869d5a55b35d2987263f75db44 100644
--- a/Simul-Robot/k002_episode_001560/sample.json
+++ b/Simul-Robot/k002_episode_001560/sample.json
@@ -222,11 +222,11 @@
   "depth/depth.zip": "91be1a0f03d0b3eb651d5b6885250f1354c8a2e06e923ef8eab694a758bf8171",
   "labels/objects.json": "31267ce5273e337421bfc9fd9a37804e5ccff308dc6cca1bf701af4b82a7205d",
   "labels/visibility_states.json": "5c58554a2b7e81efd3a230c12b9756aead32636edf8d518c970c0e1905abe53f",
-  "source/cameras.json": "c78053bd53760fd69924b1943b4c173b0a489ab2fc77ee30401515f54342301b",
+  "source/cameras.json": "f859c83c2ac40622ac55a06d2618fa2c48216eca285795c38a0a11f1f7409560",
   "source/caption.txt": "e565c0d87e038db81c0a8b33c8780cd0d46e7b068f283fed7bad1aad7166b688",
   "source/meta.json": "53c051ced9338e43ae973044dd566787329a1f6c1b44652ecf5def08fa932100",
-  "source/points_target.npz": "898ecb659fe7c718c6f1a42eb888505f8a4ea24838b43c1c8309f622d1b7dd01",
-  "source/visibility.json": "53b275abeec6cb74693ca6d28a77a922a89df18675a108a9bcfd41a6b7f41ac4",
+  "source/points_target.npz": "cbb09ce25a866f5984d6d39145c62bf297ae40d9c30175f93dc764b74c6c4f0b",
+  "source/visibility.json": "7f2b2e422ef0bbf3478b4859291981edf400b53c845715015e43870c6a9ab5f0",
   "tracks3d.npz": "bc97d8eebdf091da098cd75936501789bd1e35c9ed96fed13d96ab8823bf2dc5",
   "videos/view0.mp4": "35cddbd315cede2ffded291f71aad43d45fba5e13b839ddc0437227a2652e50a",
   "videos/view1.mp4": "8156c47dc17ba51c124b438e6c8c6af1a7cfcc962b03491567c94f39b68c6df6",
diff --git a/Simul-Robot/k002_episode_001560/source/cameras.json b/Simul-Robot/k002_episode_001560/source/cameras.json
index fb63c03ae2e409f46f9fab147879893df0a36be8..ae587570c900f83f8c7beb5f5caa3b8ba9e6cb12 100644
--- a/Simul-Robot/k002_episode_001560/source/cameras.json
+++ b/Simul-Robot/k002_episode_001560/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k002_episode_001560/source/visibility.json b/Simul-Robot/k002_episode_001560/source/visibility.json
index 9f9ab658bdbea24ec1d1bd7fa413d4ca461486fc..ac2f3bcaedf54c4362d37468ffd5a2044010d1b6 100644
--- a/Simul-Robot/k002_episode_001560/source/visibility.json
+++ b/Simul-Robot/k002_episode_001560/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "magenta_plastic_crate": [
@@ -9695,5 +9695,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k002_episode_001800/sample.json b/Simul-Robot/k002_episode_001800/sample.json
index 55ff51c450f45dbf5d6218ed219388a7b73d6ab6..47d044ce5af2a459e6216970e4a70ef2e29d9a66 100644
--- a/Simul-Robot/k002_episode_001800/sample.json
+++ b/Simul-Robot/k002_episode_001800/sample.json
@@ -222,11 +222,11 @@
   "depth/depth.zip": "295ab16c4cadd8343aa85a5dd207d6e64495f49c5e66be61d8891adacf35fa06",
   "labels/objects.json": "bb030bc0d6cf6c0b686d8232fe4ad5fb8e22cc17e12c573b664a72228d9c8527",
   "labels/visibility_states.json": "d2e0b490cefdb6311531618fe6683bac3bbd932f5371d0b1231aa58912559180",
-  "source/cameras.json": "c98d966c63770f036fba8b6e3a7cdd09e503342b57f3f0f7a51e2a6cdf55d7dd",
+  "source/cameras.json": "4c1fa5832673d3ec2e0e658267141934260f2380a1495ca5f645a3b6f5382f1f",
   "source/caption.txt": "80902c7e0cbeb6db11d3b01a6ce0f0b39c41706bca029c647a76e6c9629c0843",
   "source/meta.json": "dc2f4cefc1c078d469239aa19ffac18ce0543995d0198e48c338d9a8b8834d31",
-  "source/points_target.npz": "0a03390bd3807ad45886bec9799c47adf74954933c89222ae585b38668ce3162",
-  "source/visibility.json": "99981cd1fea09c97f765bf5cad258686a98d5472dacbf38ec549306763a381f8",
+  "source/points_target.npz": "19d79d53d0e28ee75e09b1d754b70cc0fd6c8baa35d442434df932ad248856b7",
+  "source/visibility.json": "ecf5cc6d10fb35426c9931e205accd2a20fdf56d04e985400be9fe64c94efad2",
   "tracks3d.npz": "0c28cb1d689345ad723c80c5daee084b51bf69c90cfba46fa9651eb1b7edb17d",
   "videos/view0.mp4": "2c14294c3af854827e4571cd909d5220ef76774a17d713ebf276ff74d0fea19f",
   "videos/view1.mp4": "9ef502439917435c6ec49d794f13cb3ed81210d8df3f54c00a23ed627fd93dc6",
diff --git a/Simul-Robot/k002_episode_001800/source/cameras.json b/Simul-Robot/k002_episode_001800/source/cameras.json
index 928e38c3440d9e1a6a4e3547120ddecc69ad3758..b7a76d1041f0772f066ab3d61a7471cf17130130 100644
--- a/Simul-Robot/k002_episode_001800/source/cameras.json
+++ b/Simul-Robot/k002_episode_001800/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k002_episode_001800/source/visibility.json b/Simul-Robot/k002_episode_001800/source/visibility.json
index 94ecb0c7027be4e68a66587cfa34ea7046e28d41..c1ffb8e8b1f2181c02554918f386db073b0dbfda 100644
--- a/Simul-Robot/k002_episode_001800/source/visibility.json
+++ b/Simul-Robot/k002_episode_001800/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "art_deco_bust": [
@@ -7945,5 +7945,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k002_episode_001870/sample.json b/Simul-Robot/k002_episode_001870/sample.json
index d2dfdaa0f95b683045b409502c470ea0e578d1ed..97d3b15451281f338506cd0b7cd1168277ae07d9 100644
--- a/Simul-Robot/k002_episode_001870/sample.json
+++ b/Simul-Robot/k002_episode_001870/sample.json
@@ -222,11 +222,11 @@
   "depth/depth.zip": "522feec748f66faee98a334a191d7e8501c0832471f3665691d5c1bb280d3b0d",
   "labels/objects.json": "6a264b2e78022314023db2a89eab90dffd0fcb2f5ab5d5d0813c6f778b8fc8ef",
   "labels/visibility_states.json": "b9c261351e2428005df0c1ef48b37e70be6c129a088c1fde8488262909c1010d",
-  "source/cameras.json": "7e3e6c807676daa0be23648b380b70a6c0cb8b11bcf3cd933f117f10f77835d3",
+  "source/cameras.json": "c8bd8c4552c52581cd1378e498fc383cd554ea9a76f5e3092e87cf0a3537b919",
   "source/caption.txt": "c924573ac7bd1a49eacd827959bf1abb238d25bd952a25d965a4576dbcb5daa5",
   "source/meta.json": "b361646401517f1b1b4a2526673de9f68ea76e43a77a175a70956898e1e5e03a",
-  "source/points_target.npz": "fd73d096f945d5226dfa171ae2b3b89b2f410f81e8f5ffa9082c9f00e050aac2",
-  "source/visibility.json": "6320dc4598ae0a0343d23eded304108962f401a778e42e6df21c45c7c9b14bdd",
+  "source/points_target.npz": "c3c50eaf87754e33760b3c8325ff68a42290bc6038427b6541cab8c79cf6c807",
+  "source/visibility.json": "ce3065bbafc8f2d656b5875870fcef28d25565af937dce86b4923a3a8656b5fb",
   "tracks3d.npz": "3d170875bc60d0eab8ab9505c88b021455a826e3de1796c91951e57fd3276b33",
   "videos/view0.mp4": "4e0f40f30addfa8fe86c7523a7737721be95ff3cbf71e672098ee381d5941691",
   "videos/view1.mp4": "bf62b17f8de7d5b3d6c252eb22605030ba1fb213aab20faab64e378960c88545",
diff --git a/Simul-Robot/k002_episode_001870/source/cameras.json b/Simul-Robot/k002_episode_001870/source/cameras.json
index 1ae0d07078f1e290743fa98fc6544b8baed6f93c..ba31c452e782a53257f28acb639e393bae2195e3 100644
--- a/Simul-Robot/k002_episode_001870/source/cameras.json
+++ b/Simul-Robot/k002_episode_001870/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k002_episode_001870/source/visibility.json b/Simul-Robot/k002_episode_001870/source/visibility.json
index abed86fe4e9b9f407679d3e80b2f4f1c5bd1d5ec..b45fe557e44c63d016e119587f0dafd7d245abc6 100644
--- a/Simul-Robot/k002_episode_001870/source/visibility.json
+++ b/Simul-Robot/k002_episode_001870/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "white_cat_figurine": [
@@ -9445,5 +9445,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k002_episode_001970/sample.json b/Simul-Robot/k002_episode_001970/sample.json
index 14f9ee5acf38a174e155c6debd69c6bff83db073..bbdf98f48cd95180f6fcd1372a6032aaf2171862 100644
--- a/Simul-Robot/k002_episode_001970/sample.json
+++ b/Simul-Robot/k002_episode_001970/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "0cc751972586b8856a988f4eaa114f44bec0f64c39a742cfd6a19e1781cfa50d",
   "labels/objects.json": "cb965ae860c52e46a0d183faaa2bd3cd730bd6cd5634f12c3d656e426463e785",
   "labels/visibility_states.json": "943ee8c0baab2c02007abcd1502cf592897684d960804fb02e57a593f5ff4a65",
-  "source/cameras.json": "1a2bf4226a7d284c05599bb7a18d18c6b50a8f5a8a88beb7cce2f834511c7d24",
+  "source/cameras.json": "66b283547946ae29adcbb2efe05caa556aa89ecee212f47bd63ed64686a06a51",
   "source/caption.txt": "b9129594ba25ba0fcdc7168b7669973361fb6c4eccbcd604ea7232c13b6c40c0",
-  "source/meta.json": "3dcd7733b8b0f61946eccb806958e6940f52eeada35b1192651031074fe876ef",
-  "source/points_target.npz": "3e69d6e84a491bb4c6a6298d43ea154f01f2df097de798f732efb827a2eea0a3",
-  "source/visibility.json": "f930c90ce6b7821a09eb66762f5743edafa11e16fac3cba59de676d334261aea",
+  "source/meta.json": "76db52bf3f8af2149b146fc934346e4de2f225735acec3279a268a35154d0163",
+  "source/points_target.npz": "264a044af08f50f6a4439adc3fb758bb5839b80bce6e4ba9e3edb5e639e9826c",
+  "source/visibility.json": "fdb9719256ac38b0f2fd1a04ccf020f2d108083ddd726816bf2c190bd2432d23",
   "tracks3d.npz": "4a638f4eabbaf3ace9bd636633651a88f25e81ce5421fc732235e4b9882a5d7e",
   "videos/view0.mp4": "46ac7204ab1ba00ca1f41cb015e0119ca78080cb4370906f2911bc22a9a395fe",
   "videos/view1.mp4": "1ad374b6b9ef87fea04bce0625095cb0a8f53521ff1f41e850c02cbb0e7a8233",
diff --git a/Simul-Robot/k002_episode_001970/source/cameras.json b/Simul-Robot/k002_episode_001970/source/cameras.json
index 36c306c98fb2c048a498b80eb4d0268f39378683..bc6847a5ae676956097c8ded5ba3b9dd4a0e50aa 100644
--- a/Simul-Robot/k002_episode_001970/source/cameras.json
+++ b/Simul-Robot/k002_episode_001970/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k002_episode_001970/source/meta.json b/Simul-Robot/k002_episode_001970/source/meta.json
index b0381fc1eca0d7ff8407261727b87f32fcd59f26..dfb7c9f9f4ad445fe65a53cdff0de61e31a00cd7 100644
--- a/Simul-Robot/k002_episode_001970/source/meta.json
+++ b/Simul-Robot/k002_episode_001970/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/k002_episode_001970/source/visibility.json b/Simul-Robot/k002_episode_001970/source/visibility.json
index b681f70dbf03f9bddd21907a2745f263be977216..ac3ddbc64c442bf8d5cd825c8b357a3a63f66eaf 100644
--- a/Simul-Robot/k002_episode_001970/source/visibility.json
+++ b/Simul-Robot/k002_episode_001970/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "cream_ceramic_cat": [
@@ -6010,9 +6010,9 @@
       "target": "cream_ceramic_cat"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -10000,5 +10000,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k002_episode_001990/sample.json b/Simul-Robot/k002_episode_001990/sample.json
index e5731063c28a7f20f1cc18d9309d8b7102646182..816fdfe8914e20ea620a99d3836d5a9e4e1cd50b 100644
--- a/Simul-Robot/k002_episode_001990/sample.json
+++ b/Simul-Robot/k002_episode_001990/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "5a5e5d13de4d5ee2807881235c88efe0a32995fd08519852bf8ad3d4b0be3243",
   "labels/objects.json": "d61dffeee6e5d4dc95dc75cfb6d97fb772e4b9ca68046d838480e3d1918ec620",
   "labels/visibility_states.json": "e3913048232501e467ae9e781fabd0fd524da3412b29256752853280f46b4fc2",
-  "source/cameras.json": "3f3877af46a91473009230d7873e5818a6d725ad5e5d0b000407417975539d68",
+  "source/cameras.json": "5b9d31f3ca098e645ff5feeaeb1caf4e9f50cc7ec0d99dab27eae1a6d3235b7e",
   "source/caption.txt": "5e2b9055f9d005109d0737bbee0355351970224800d05e373fd0c55edfedf39e",
-  "source/meta.json": "8754a3a664956c97b0753968e92edcfa8a7c5f7e922080f45d1d9e6a99bbdc32",
-  "source/points_target.npz": "73298fae754aef8f8b048a7827c25f311bb59a3f95d09aee1512d2dc89d1c0d3",
-  "source/visibility.json": "c446e9b28d81e0d33599ca9ec88cbac48afcf45632e4fd6162219cf054bd9b60",
+  "source/meta.json": "d8fe8f236c7d131a8ce98f594f33ea9d6c4749476aa7d03eb948388c42bbfabc",
+  "source/points_target.npz": "316eeb4d2521af92531afe36345372eb13c19f8cebf2d4d90d7ad0796871d8c1",
+  "source/visibility.json": "82bf5e8dc01191cc53ae12fa80b7c108da79d03440cdf7acefaf7c5ab6888e43",
   "tracks3d.npz": "75bea202dcaab0167339fb16c773aac2a11aea71e34435144f33030fc4baf1b8",
   "videos/view0.mp4": "a14c711a25e5531f062978d80bf7d4de3d5edffd5670cc8be304fc4632124ea1",
   "videos/view1.mp4": "2ae0068d37fe271255eeabd77261268a92a3b63a81f526ffcf5d56b203fdfcdb",
diff --git a/Simul-Robot/k002_episode_001990/source/cameras.json b/Simul-Robot/k002_episode_001990/source/cameras.json
index 8379accda70d3f7edceeabae0fe9ddd5ad1b8985..49976099d7010b925a6a4b152676bf7a4d25139e 100644
--- a/Simul-Robot/k002_episode_001990/source/cameras.json
+++ b/Simul-Robot/k002_episode_001990/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k002_episode_001990/source/meta.json b/Simul-Robot/k002_episode_001990/source/meta.json
index 7c37d9bc4ad4993886e1eeecc79e33ed1e841dde..deb7a7b64274f27822c12975764f5b22ef8fdca3 100644
--- a/Simul-Robot/k002_episode_001990/source/meta.json
+++ b/Simul-Robot/k002_episode_001990/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/k002_episode_001990/source/visibility.json b/Simul-Robot/k002_episode_001990/source/visibility.json
index 797eaca7356a05fc242f03d20377c5a640488567..cdf01e3308c809b56e8012049b5a7fe9a7d462d0 100644
--- a/Simul-Robot/k002_episode_001990/source/visibility.json
+++ b/Simul-Robot/k002_episode_001990/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "amber_cat_figurine": [
@@ -6910,9 +6910,9 @@
       "target": "amber_cat_figurine"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -11500,5 +11500,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k002_episode_002040/sample.json b/Simul-Robot/k002_episode_002040/sample.json
index 85eb4105ece9036c31aac947cd80626839515d9e..5cf568b1d4e82628aff9efceb5f54ec936041b84 100644
--- a/Simul-Robot/k002_episode_002040/sample.json
+++ b/Simul-Robot/k002_episode_002040/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "170caaa40227a56d8a7f15700a8b6564c408305d7486dace68b2dfbf95d56405",
   "labels/objects.json": "71a5088b683ffcd6641fade2bd337520cc6f840149307ab915deebc90b9fd19b",
   "labels/visibility_states.json": "ab91a46ae25af37ad71cfd73229e858c3023a789c69511d1a76d3326eb2a263b",
-  "source/cameras.json": "0564ed9dd86162ae6b643da623ca01bedf97d1eba9f97b4eba5f5acc115e2f59",
+  "source/cameras.json": "402f4e4ca719cb6273a8d8907116980b3655992740bf442f9ce775cea3b5b8c4",
   "source/caption.txt": "bb2c1932980187c3a0fd1d3121ad33ed36c999ef67731dfd2b6a3c16ff06fbad",
-  "source/meta.json": "193fd50a4a88ad4cedf1982b7d5844053fcf5baa8c61f7ec7decaef3fe93908c",
-  "source/points_target.npz": "dd33f6f82ce08044649743db9accaf46150e4fee73ea3ec0a8c7a3db74fdedc7",
-  "source/visibility.json": "9e7b2914ad23a99a54475a7313cd1280b32937519da02c5394395195e17bd3fa",
+  "source/meta.json": "453ac938f05ec87dfc451ff5f5af88e0871550ed234268e9dea8c4339cb7b8cc",
+  "source/points_target.npz": "f3b972791462016be24750b5c7a1447182508c6fd1f41728def185fecc509977",
+  "source/visibility.json": "ce29a35adfa42d2b6be5ec8a06ab9d2cc52931beee51352439aed312a3f07094",
   "tracks3d.npz": "3cb8fed542d3e64ef3516e389e5a0693d8455c01917e07c17c857326766e9f0a",
   "videos/view0.mp4": "c0d9f25c36ef5846bc5049c79a11905c56054a8e601baa3fdc61eb8ee5de5d0a",
   "videos/view1.mp4": "becc9b3f247f9a582ced46f509ef6782313ffdd00e0aa938c72ccba056ca581d",
diff --git a/Simul-Robot/k002_episode_002040/source/cameras.json b/Simul-Robot/k002_episode_002040/source/cameras.json
index f2e0898f054b4c96cc494adbd8b50905876d243e..b723b9167b24806d7e907b638c09e9a860b9c002 100644
--- a/Simul-Robot/k002_episode_002040/source/cameras.json
+++ b/Simul-Robot/k002_episode_002040/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k002_episode_002040/source/meta.json b/Simul-Robot/k002_episode_002040/source/meta.json
index 2963cab8053b19800d692d57de75c25f230d5d85..2b297b4a9cca00e70811946e0d5c72b7d433a790 100644
--- a/Simul-Robot/k002_episode_002040/source/meta.json
+++ b/Simul-Robot/k002_episode_002040/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/k002_episode_002040/source/visibility.json b/Simul-Robot/k002_episode_002040/source/visibility.json
index 00ac71c84834171250d0d78f3a3bf45d5f69cfb4..70aa35c9a26c557cde0174c0ff5e6c63268881ad 100644
--- a/Simul-Robot/k002_episode_002040/source/visibility.json
+++ b/Simul-Robot/k002_episode_002040/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "square_big_box": [
@@ -5680,9 +5680,9 @@
       "target": "tomato_soup_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -9450,5 +9450,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k002_episode_002080/sample.json b/Simul-Robot/k002_episode_002080/sample.json
index 74365e3e918c62bf46b59cd946c49dada7fcbb70..8dc881c015c9661e070992cf3e1cc20fa3dc3a96 100644
--- a/Simul-Robot/k002_episode_002080/sample.json
+++ b/Simul-Robot/k002_episode_002080/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "0f353b14cf5daa2146f90407ea17888c61ba2b5a8819b3a0543904a805fc43e6",
   "labels/objects.json": "4f5544862ab107dba47ba3cd47e9ea988ca445b748b8a85528dbc652bf48637f",
   "labels/visibility_states.json": "43815368dc55be7c59631086ffc66392b6ca3ba25fc602b8e7383ce415905856",
-  "source/cameras.json": "680abd416e03795baa050238e4e2ece31c8fb201f518ff1ade88588acc413d4f",
+  "source/cameras.json": "ec0fd06fae1f882389a0907d9a9117d6d3be66b3d18ef6a9d9d0dbf76e13af3f",
   "source/caption.txt": "8a6e17d8477cf7ba67064f9b8e55305a1bf8a24541faf2543d8c23d1e984621f",
-  "source/meta.json": "4171f8b2ec869eb9d0c810954838ce50a6a2914e683eb16d50f7b735df66ba82",
-  "source/points_target.npz": "5296cb9fb4f3c4b3c418ef2c805871f3e6b48f31dc50f33cb27fa7afd8d3b8c1",
-  "source/visibility.json": "27bffc9ad9a8db8f6cbea9ceff2ced0af5f232181c69a11c351420d2f600abf2",
+  "source/meta.json": "e7c9873eabb31cd823a756ad36f1187c0dad67c479edd51163f9b2a7dc2c5d71",
+  "source/points_target.npz": "7e0d59afebc909735f5b1dd937b9de975b5544b455c1a9ccb9d2fe5994003bc1",
+  "source/visibility.json": "c288483e6eafc03bb30fc353604490877a1bc967bd73bb336f7a5450093a87c0",
   "tracks3d.npz": "525a0a94d4873e92628f08fce5053a782bd9f982a3543edbedc8d7888910da05",
   "videos/view0.mp4": "85637c8ff7f3d97ab3060abeceea63d8a9286e33ef93d33effc049c218c5691a",
   "videos/view1.mp4": "ddf90196104a3ecfa9b880d4148e3a2046ac75cd19b45af659e5c613344e77a6",
diff --git a/Simul-Robot/k002_episode_002080/source/cameras.json b/Simul-Robot/k002_episode_002080/source/cameras.json
index f67cc8a695a40c809aa3462c109b9a6fb601a0c0..8a568382f7ff5436166adeb8e00fabda403c20a4 100644
--- a/Simul-Robot/k002_episode_002080/source/cameras.json
+++ b/Simul-Robot/k002_episode_002080/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k002_episode_002080/source/meta.json b/Simul-Robot/k002_episode_002080/source/meta.json
index b26ad4e6b444084a7337e0f401755cacb6bf816d..d392f91f37fee0293d0a7e276d671be4f7f82a31 100644
--- a/Simul-Robot/k002_episode_002080/source/meta.json
+++ b/Simul-Robot/k002_episode_002080/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 0,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/k002_episode_002080/source/visibility.json b/Simul-Robot/k002_episode_002080/source/visibility.json
index 9c78d7654a5142e281fd847a5525c04de443b63e..11a883793298372845c0a11c3b73bba979d41a16 100644
--- a/Simul-Robot/k002_episode_002080/source/visibility.json
+++ b/Simul-Robot/k002_episode_002080/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "red_milk_carton": [
@@ -7810,9 +7810,9 @@
       "target": "red_milk_carton"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [0, 0, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, None, None] < 5.0cm (못 봤지만 잃은 것도 없다)",
-      "④3캠 합산 occluded 0f < 20f — 카메라 회전(out_of_frustum)만으로는 이 조건을 못 채운다. 팔·디스트렉터의 실제 가림이 필요하다"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [0, 0, 0] f < 30 f (a single camera sees the target over the whole clip)",
+      "③ target displacement during the invisible runs per camera ['cam00', 'cam01', 'cam02']: [None, None, None] < 5.0 cm (not seen, but nothing was missed)",
+      "④ occluded frames summed over the 3 cameras: 0 f < 20 f — camera rotation (out_of_frustum) alone cannot meet this condition; real occlusion by the arms or distractors is needed"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -13000,5 +13000,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file
diff --git a/Simul-Robot/k002_episode_002100/sample.json b/Simul-Robot/k002_episode_002100/sample.json
index b822684a4022f644d689e0cc06e20aaaa389e137..018aca12098a8d33db266df340d5a77c03f09435 100644
--- a/Simul-Robot/k002_episode_002100/sample.json
+++ b/Simul-Robot/k002_episode_002100/sample.json
@@ -222,11 +222,11 @@
   "depth/depth.zip": "f317aa3197a6e992342d92727481680278e6f803b93e7bdab959bc1f511d07de",
   "labels/objects.json": "818841f4c27397226cca6bf8eb770fd833b0e2a7c47f09e9dd9146ae76b30ac1",
   "labels/visibility_states.json": "e0008ad420dba57c9d482063a5414ba32cd20ce4166b6b40b13021091862975e",
-  "source/cameras.json": "36667fde7c6d95c20e18959d0d08c3480f7a4163d448364b755f4a458f1a2a07",
+  "source/cameras.json": "18f384278ef29d7927caff5e0551c063c4ebb094de75ea30d13525a99e78abeb",
   "source/caption.txt": "b5a85bad7b862200af2efb907ea2938bca6a64026aa53464b1329ffdcd18ee26",
   "source/meta.json": "e1532a8dd86c327ac1ecc33c5eeaebd5a11613570bcbedd431f0eba1d851bbad",
-  "source/points_target.npz": "45554fc0dbc2f653baf3bd9036d12a0b299379f84c6df040a517ab9481be5606",
-  "source/visibility.json": "0e99cd62d54971308c72a0a87f6ef961b4a82ddd7961808a5eb0ff50ea0828ed",
+  "source/points_target.npz": "c0a6b178102f10b4f2b62b2bed560ed775a67f740984d16c64ccc7e2af3d8f93",
+  "source/visibility.json": "7540dfc058294a5451ae4f8288c72797014b0096c5c708ff3fe4c8b6ec1dbff9",
   "tracks3d.npz": "4069771af08e8a390376c9b4cc0673caec092cc78372e6ae1faadd4b090baf73",
   "videos/view0.mp4": "1c3a752e24b07503f753b39eeb577015f81b0eae5a5bd9bfc588815c2afbb4a6",
   "videos/view1.mp4": "216eaee924371a05e6d6acbfe6c80360b67ba55536cfc62a10d67fb32f3c2802",
diff --git a/Simul-Robot/k002_episode_002100/source/cameras.json b/Simul-Robot/k002_episode_002100/source/cameras.json
index c9cdd0ae6823b5cd184c5d9f939cca1f6b4f307c..4197c75d0dc400d04ae4d40710c5da4b8d135beb 100644
--- a/Simul-Robot/k002_episode_002100/source/cameras.json
+++ b/Simul-Robot/k002_episode_002100/source/cameras.json
@@ -14,7 +14,7 @@
     "cam01": "pnpcam01",
     "cam02": "pnpcam02"
   },
-  "note": "모든 카메라가 `w2c_frames`(프레임별)를 갖는다. 카메라는 로봇 base_link 자식이라 시퀀스 키가 부모 상대이고, 여기 값은 PIE 에서 되읽은 **월드** 포즈다. `w2c_frame0` 은 참고용이다.",
+  "note": "Every camera has `w2c_frames` (per frame). The cameras are children of the robot base_link, so their sequence keys are relative to the parent; the values here are **world** poses read back in PIE (Play In Editor). `w2c_frame0` is for reference only.",
   "cameras_data": {
     "cam00": {
       "name": "cam00",
diff --git a/Simul-Robot/k002_episode_002100/source/visibility.json b/Simul-Robot/k002_episode_002100/source/visibility.json
index 74fae7605a19d57b9edb5b69f506a6b8053a9df2..57cd9b938ea5439a09ded00a406267a114486338 100644
--- a/Simul-Robot/k002_episode_002100/source/visibility.json
+++ b/Simul-Robot/k002_episode_002100/source/visibility.json
@@ -36,7 +36,7 @@
   "states_rule": {
     "occluded_visible_frac_max": 0.1,
     "denominator": "points inside frustum",
-    "note": "visible ⟺ (프러스텀 안 표면점 중 보이는 비율) > 임계. 임계 0.0 은 '한 점이라도 보이면 visible' 과 같다."
+    "note": "visible ⟺ (share of the in-frustum surface points that are visible) > threshold. A threshold of 0.0 is the same as 'visible if at least one point is visible'."
   },
   "visibility": {
     "microwave_popcorn_box": [
@@ -8545,5 +8545,5 @@
       1.0
     ]
   ],
-  "note": "가시성은 렌더된 16bit depth 대조로 판정했다(ε=3.0cm). 카메라 3대 모두 프레임별 w2c 를 썼다. npz `vis_cam`/`vis_cam_robot` 과 이 필드는 같은 계산의 두 출력이다."
+  "note": "Visibility was decided by a depth test against the rendered 16-bit depth (ε = 3.0 cm). All 3 cameras use per-frame w2c. The npz `vis_cam` / `vis_cam_robot` and this field are two outputs of the same computation."
 }
\ No newline at end of file