diff --git a/Simul-Robot/c001_episode_002590/sample.json b/Simul-Robot/c001_episode_002590/sample.json
index 74d408993f7eebcedfcc96d585fdb91e99b2f621..4c6a006315cb52f4a5de7da5e1b675b2438b080d 100644
--- a/Simul-Robot/c001_episode_002590/sample.json
+++ b/Simul-Robot/c001_episode_002590/sample.json
@@ -222,11 +222,11 @@
   "depth/depth.zip": "8f57e12e842bb1122632249d67dadfdfa3cc9bdca08bf07a57afdd7c73096141",
   "labels/objects.json": "6d94f5ac0b0d351d06f0c42fb60dbc068c6ae7c3991d942324ae8b2201a9847b",
   "labels/visibility_states.json": "90cb4085f5425b98c06bfb80cb4d7c0362016482e9deb8474664247a2744cb30",
-  "source/cameras.json": "7818881712128f575efcdcd076ed8537bd40a0d036c107b5e1d0869f24dcc4c6",
+  "source/cameras.json": "f57e5dc45c7fd08b569167cee6edc676f62d57420fc340a6afb09ebfc312f299",
   "source/caption.txt": "319eb974f7064e8f2a6fc034fd0b834ec784f4184d70252fe751344d039791d5",
   "source/meta.json": "6e9cf0ee2419f553dd99dd368fd4aee970e4b798692220099e63f8901cd56cab",
-  "source/points_target.npz": "2ce97407d42792d87b42f3b7cdc95c0da7ac5092e9fd05f20493dece5e8eadc2",
-  "source/visibility.json": "1fdb4eed68a08f44c8a97032eabfdaf70a0e80594b93750c5d9496eacea6c295",
+  "source/points_target.npz": "6e7df7fccc850989b4098db362cedd6139b30e742240208958e9cecd04633acb",
+  "source/visibility.json": "a09134a8cacaa721ae9f020547c82808e42ac6891b64e5462ef4753f553d7165",
   "tracks3d.npz": "87c6f53fd705d5ef452c65101e76c7f55a35e0e70c08ca7eff0cb99cc6b7a657",
   "videos/view0.mp4": "6c0fe9d8d99080205b30ff8865b9451e7e2c1697553d4e3f38254e004d883d88",
   "videos/view1.mp4": "16bb71e6b57077af2b80e027e2079a214c7693dfa57449ed65cee900e240ecfb",
diff --git a/Simul-Robot/c001_episode_002590/source/cameras.json b/Simul-Robot/c001_episode_002590/source/cameras.json
index c7e26c520657b4382741d3d1da7600cd6db1f0f6..cf49a482ab85f853de5aaf5e34b629b7e2e34cf7 100644
--- a/Simul-Robot/c001_episode_002590/source/cameras.json
+++ b/Simul-Robot/c001_episode_002590/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_002590/source/visibility.json b/Simul-Robot/c001_episode_002590/source/visibility.json
index 14b155794ec2b991ace0ac9870253f383cbe4d4a..7b1d3e9f66b6c45dc9c457a4504432b569697ad1 100644
--- a/Simul-Robot/c001_episode_002590/source/visibility.json
+++ b/Simul-Robot/c001_episode_002590/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": {
     "kraft_coffee_bag": [
@@ -9345,5 +9345,5 @@
       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_002660/sample.json b/Simul-Robot/c001_episode_002660/sample.json
index 538b93d8b3e4ac3bb0d9ab1f4107167e21f7ecd4..2be391fea38932d829470f48bbef6c23d7b7c1fc 100644
--- a/Simul-Robot/c001_episode_002660/sample.json
+++ b/Simul-Robot/c001_episode_002660/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "3c3d2195cdb225ea20c70bc57599edaa342d3308816c1875de3d844d6aedc67c",
   "labels/objects.json": "98eb85444f0622cc84215024b2f0d6c491a6296610f978579d0c56c28bb1bdaa",
   "labels/visibility_states.json": "ac7fd963280de9c673db015c964450195b08b6043b94dfd4a55c2fbae32bbb3d",
-  "source/cameras.json": "a2acd227b4567fda20230f7b89bfbfafe397a465794b84f710080f33d78c725f",
+  "source/cameras.json": "3154ea25c85afa15d06b170ce274353720e1c6fd6bf98831c88321d905536504",
   "source/caption.txt": "27860719a1ddec0bdab6feb6189c77ca15d48e9c025f0066de81e7adb1dd0a70",
-  "source/meta.json": "152ca58457d5db30096530226710e21b6604bd11e63f9777ef7a1987b65d8003",
-  "source/points_target.npz": "077c3c5558c1e8823713ee48632bc15196ef2a744394f61bc1554b5a23235b7c",
-  "source/visibility.json": "b1e64bdb600c336d3707b359873e3eb11a8076d561bddbb4d01723cd7fd4be11",
+  "source/meta.json": "29ac371253e0cea5a85afe8f5d413eddc05937c780dd618b6934af75cb7461fc",
+  "source/points_target.npz": "5a3d8b307ae3f007a0564a8ff5b8fcf7e6de7c99b69b70844248a52fc561df62",
+  "source/visibility.json": "92807b4493eae9e311fb94a40e88f6553446d562d25b4f2bd10f3a4f7e1f975c",
   "tracks3d.npz": "fcdb389638eba6ebd8ca4239d4ccd5b23e6a1cf57fb248dfe7ec5e3bd2a4bdd9",
   "videos/view0.mp4": "06731f6e8b45ecaa6c97a1d0ba6ebabb675fdf9872275b13e2ba4591d62726a8",
   "videos/view1.mp4": "537e06def5cf0235f0a320d19096e5e7bccfb087e2b5d27d75825f59de4f9bfb",
diff --git a/Simul-Robot/c001_episode_002660/source/cameras.json b/Simul-Robot/c001_episode_002660/source/cameras.json
index c4b62b765ed06bae77dfec8adfd5011873a531ca..10c9c995002357e75c437d4925988b6d013b3a06 100644
--- a/Simul-Robot/c001_episode_002660/source/cameras.json
+++ b/Simul-Robot/c001_episode_002660/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_002660/source/meta.json b/Simul-Robot/c001_episode_002660/source/meta.json
index bce9bf441c22b1fd128fdf8bd53f6d6aebbbc3c3..32d6ebcafd0d8044ce71add955023828182de1f2 100644
--- a/Simul-Robot/c001_episode_002660/source/meta.json
+++ b/Simul-Robot/c001_episode_002660/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_002660/source/visibility.json b/Simul-Robot/c001_episode_002660/source/visibility.json
index 73cfcb8aa786117432e8f75bd97c8817e12b5acc..03f97eb4aaf26f75b2b6b45d8030c3927c1b7402 100644
--- a/Simul-Robot/c001_episode_002660/source/visibility.json
+++ b/Simul-Robot/c001_episode_002660/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": [
@@ -6610,9 +6610,9 @@
       "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,
@@ -11000,5 +11000,5 @@
       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_002760/sample.json b/Simul-Robot/c001_episode_002760/sample.json
index 1f7c49c7811571d27587a3653c57534bbdacf385..ee68a3ac731529416570af1d88dc49b9f24407a6 100644
--- a/Simul-Robot/c001_episode_002760/sample.json
+++ b/Simul-Robot/c001_episode_002760/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "28bb25ed7cf31f887873a0e08aad59f87124ed0590e9d565780e709e5536366c",
   "labels/objects.json": "4d580d667df69397d6952b73c9b971da2ef71d019f2af4c9e9dbf8eff3b45756",
   "labels/visibility_states.json": "0e68715ffd883678e19f8d3daa62c9667a61925502adc8d411be7ecade6dd100",
-  "source/cameras.json": "f46782b84c5cfff7417c519ac7a16620459a18356e6929dac76949b691c270c5",
+  "source/cameras.json": "4811cbb681c42beff68519b0b33ae5c40a84a87be51949ca043244b92928f557",
   "source/caption.txt": "4fc336679d6a860b4030ccc81de0f0871ba460e7858d4e437c3deb3a29d917e9",
-  "source/meta.json": "344d98b240d80699d1b1fb4ad3722fb7b9bf2b8f58c000badabb7b0230516880",
-  "source/points_target.npz": "70192a5dc695e4bf1107f8cb57e747ecc426a1418325aa7aec2fbf64bd8da44b",
-  "source/visibility.json": "acfd6d38ec312b033519b30c47f33b92afac89b096e8aee92478f8198204c511",
+  "source/meta.json": "7c66576c3bb3d5df74058f0ea33b4f43f8e4fd74451b683ccb8f0ffdb675d174",
+  "source/points_target.npz": "074d94c75e276609e352cb7b183f53828b3394e352d01915fca6e860b16dac9f",
+  "source/visibility.json": "c7a0db4542c94a567da1ae216210075eb50b73a9288785a259518c07d1e18773",
   "tracks3d.npz": "c763ad5838c20e5cfac7d8f2467b2436025ffd67800c0365248cf43deb669a0d",
   "videos/view0.mp4": "ce17e62a427ef08b915ad6eb74eaa13d4bae90547f27277ed4b17004a433cfd5",
   "videos/view1.mp4": "f75dcc39295786b516fbb4e11cb7c65a382c8d40c20a0a9bf4dea15853efc831",
diff --git a/Simul-Robot/c001_episode_002760/source/cameras.json b/Simul-Robot/c001_episode_002760/source/cameras.json
index 800a9027e56d1e8b1d2efb229a62eec80563da3e..7cc0f890fdf5ce10eb23479860ef5ebf0e7b5e4d 100644
--- a/Simul-Robot/c001_episode_002760/source/cameras.json
+++ b/Simul-Robot/c001_episode_002760/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_002760/source/meta.json b/Simul-Robot/c001_episode_002760/source/meta.json
index 999db645e517b7851cc2b58f28d05d1c5ff83604..471b0a3f642474442be27dba8bc9b6da66d0d163 100644
--- a/Simul-Robot/c001_episode_002760/source/meta.json
+++ b/Simul-Robot/c001_episode_002760/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, 1.56] < 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, 1.56] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_002760/source/visibility.json b/Simul-Robot/c001_episode_002760/source/visibility.json
index f88d08105225bfb8f59da26a0b2ba7be1cf70ca0..5b408ff0f8b71da165d59c30bd8e9e0d7e668095 100644
--- a/Simul-Robot/c001_episode_002760/source/visibility.json
+++ b/Simul-Robot/c001_episode_002760/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": [
@@ -6883,8 +6883,8 @@
       "target": "blue_tuna_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, 1.56] < 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, 1.56] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -11452,5 +11452,5 @@
       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_003020/sample.json b/Simul-Robot/c001_episode_003020/sample.json
index 26729a710d660b3b096cc5345fc8ae2daff2be78..e49df9a3830854f9746fbe2d0c823760d8031fcf 100644
--- a/Simul-Robot/c001_episode_003020/sample.json
+++ b/Simul-Robot/c001_episode_003020/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "1d709ebb250b95b09a0219cf26530abbc3e7c70cebb9357daa266ce396d595cc",
   "labels/objects.json": "734640d61bf52ac73c60b7c32b0d1c74335e72f58463281d7a4e08df879f763e",
   "labels/visibility_states.json": "52f115262fcc285b8a3716fd7efeb3d448a3fd7486901767f1ba8615a73c6b3a",
-  "source/cameras.json": "2f481ac82ad16032a1f3f8954c67e20c4728c11778bb7a9c49d63f96a189e24b",
+  "source/cameras.json": "af16357f22c5fd2ee706bce796e0acfd9c6fafc40baf8389f1ffcfb17afbc48a",
   "source/caption.txt": "7b94d0c3300d5651ea94183cf5d29e9eabf0b7cd3f05e0eea7c10bc941052d4e",
-  "source/meta.json": "5ee7dcbb477cdb2834b10e9cfd78b95cb84bef255d576fdb324ca28f19224488",
-  "source/points_target.npz": "16ac95977d5ce333dc4c8cf30de1084d02e318611bd44763d2ab7686349c8cf2",
-  "source/visibility.json": "97541b89f8ff8f22908a5e7aafb64a0887695d91f36ba9a4bc87254d1afe36fd",
+  "source/meta.json": "79faae92574a1fe766e14f6ad65c73573206232dbc00946dea83f5df49a2a289",
+  "source/points_target.npz": "b556bb31cfdb7215f9610eeaf0fa1987af1f415c65c552590086d25471ed798c",
+  "source/visibility.json": "d9f9f403549dfbb013153007e9a143fec95b5e3353c0f00a4542b86ab6227ac5",
   "tracks3d.npz": "ee31cf44906c9775703a28c5cdc3b3f8cab549a1a6c668d8d0d4ffbbe8cb29a8",
   "videos/view0.mp4": "c66e8420cc11b4d798b2a9c25113beffb1c96232efee6b8f3fe0e318d9937d30",
   "videos/view1.mp4": "437ed78fe5905161530f0393b5d7734033d587f2f41a604d811fd0100f20d896",
diff --git a/Simul-Robot/c001_episode_003020/source/meta.json b/Simul-Robot/c001_episode_003020/source/meta.json
index e944582521c91e172b8bf178085c4a379adb31b2..dfe09cc65f2a5685eefa78a75203ea5306e8d80f 100644
--- a/Simul-Robot/c001_episode_003020/source/meta.json
+++ b/Simul-Robot/c001_episode_003020/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 [0.1, 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']: [0.1, 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_003020/source/visibility.json b/Simul-Robot/c001_episode_003020/source/visibility.json
index adc524a7da3b19063ee2cff025998e93122d05e0..4fdf246d3e1a55cbb2b2ff3cf36c2b1d9d97f1f5 100644
--- a/Simul-Robot/c001_episode_003020/source/visibility.json
+++ b/Simul-Robot/c001_episode_003020/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": {
     "alphabet_soup_can": [
@@ -7633,9 +7633,9 @@
       "target": "alphabet_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 [0.1, 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']: [0.1, 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,
@@ -12703,5 +12703,5 @@
       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_004150/sample.json b/Simul-Robot/c001_episode_004150/sample.json
index 571c2ac7f716966a75b612689e6269e04fb763be..9bdfcf6112532839e694b7a76cbcc22290919a3c 100644
--- a/Simul-Robot/c001_episode_004150/sample.json
+++ b/Simul-Robot/c001_episode_004150/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "a25521c86f3179b3a2ef506c10f9c1ae54bfee9b25796cfb27d8b3596acbd6a1",
   "labels/objects.json": "d3c3ddaaf196ce82081c1e9995e322d70b499324000c2c1b1ecca16d74181d83",
   "labels/visibility_states.json": "21ed55f47d3a0d75ea37e01eebf163721f92199ad18087dc531d159bc1436043",
-  "source/cameras.json": "c0fce344182870d80420c9aed361792f1f0d1bec012859eeef11f5fa92cad50d",
+  "source/cameras.json": "0a53eb09a2d3d6c93983d2a755bddb4d308c118a110455990a48216249344497",
   "source/caption.txt": "9a56ce81c9ad4e6ae678baa051bd3db6b6cca8a8341f525c3192a8fcf715aaee",
-  "source/meta.json": "5c02e6e2829c1eec1109da9cf9be9eb79264fd72879d4ced4718287c7714846f",
-  "source/points_target.npz": "ddfa411d8b27971f519bed3500965675abb5b06ed86ccbf28f42002bc78eaf07",
-  "source/visibility.json": "6b7b30e742bdc7e16a8083c237e3ac0d53b19571fa7db2276b5dcfb74ddea0df",
+  "source/meta.json": "73664a16265e3e54dbaf3ba38bdf816407d6f420d0d53cab7698f40912ca8114",
+  "source/points_target.npz": "2fe9a92789b4633ee0c6437812a712076dfd3eef0b77314145a923367e2b86e5",
+  "source/visibility.json": "dcfcd8967c5d8595a73baa55f6d8da2c13acf6228eabbee17bdbcb160003cdc0",
   "tracks3d.npz": "4590a5719228fde91cf4fa6ff5fbf5053460516c5404659fc442479ba0721da4",
   "videos/view0.mp4": "a6b8d0998906f0d1ae282f9b668bd03996c89b45825a265f5ce7f646281502ed",
   "videos/view1.mp4": "e4c0a1036dc67b0d074c9ab52d0634a7e935b187f628dc874ca4377e7ca82bed",
diff --git a/Simul-Robot/c001_episode_004150/source/cameras.json b/Simul-Robot/c001_episode_004150/source/cameras.json
index 77142a38444bb6cd5f3d8aa8b2d3680e20a61815..3f09e6c11802051f1748211c379caabd2e7651ff 100644
--- a/Simul-Robot/c001_episode_004150/source/cameras.json
+++ b/Simul-Robot/c001_episode_004150/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_004150/source/meta.json b/Simul-Robot/c001_episode_004150/source/meta.json
index b7ffed0c4c2e272fc8a8854cefad85893d798460..552649a7e1dcf85824209653391524de6da21b49 100644
--- a/Simul-Robot/c001_episode_004150/source/meta.json
+++ b/Simul-Robot/c001_episode_004150/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_004150/source/visibility.json b/Simul-Robot/c001_episode_004150/source/visibility.json
index 078a2412f6a7b636f52f0d2a664555ccd516b2d9..f98769884eb028cc71ebf857389efe38324de912 100644
--- a/Simul-Robot/c001_episode_004150/source/visibility.json
+++ b/Simul-Robot/c001_episode_004150/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": [
@@ -6550,9 +6550,9 @@
       "target": "potted_meat_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,
@@ -10900,5 +10900,5 @@
       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_004660/sample.json b/Simul-Robot/c001_episode_004660/sample.json
index 0c3f9840ffdc87dac79192e45e7cae5bbfe29d8b..8e41ff8708ad23e95f247ae6d80585b2cab9950e 100644
--- a/Simul-Robot/c001_episode_004660/sample.json
+++ b/Simul-Robot/c001_episode_004660/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "c14c1b95e3050e1d06a22c20c32ab5c3d6e6ef0ae0bb60884a93d4b6633f65b8",
   "labels/objects.json": "aee0e58ca8c51928007e7bd7e1e91178bc485d24aacd2e9850fce4a80995f315",
   "labels/visibility_states.json": "d74ec6f831812140ad8089f6dea5634ef68951282b80603de805e0e82654e5ce",
-  "source/cameras.json": "dda30915ba11c6c6660075fbb48776e941a92da6274bf4817e200534bcd38a43",
+  "source/cameras.json": "8baaa7621dfdf3f574155e21ff14bfe711d375454253db97b6f26001fb3394c3",
   "source/caption.txt": "b7670a5784f50a48c68226d2f4721fee40806aaa47cc60db7097ff24dfd63eff",
-  "source/meta.json": "0d379dcfbe4a1a7ddab8185a9f02bfc4bb305059061a8d19a258c761ada72c39",
-  "source/points_target.npz": "23a9fb8899c089e1ad09da0d98667a4e593b4c0f0a734a0f4492a9379a5212d6",
-  "source/visibility.json": "262e6bcbd9c89a9a670390f95db5aa3bd7e05e18d57f53239c6769c0cef10601",
+  "source/meta.json": "d34e9363fb4934d6b618429ee579df84b526520474d2adf6af8abadb36179df7",
+  "source/points_target.npz": "3b4a9a3eef2a41d6ed022328959ddbf0b30c733108de3b0bf9a7f50dca2f8796",
+  "source/visibility.json": "791fb90ae30ddb1a15eea1600fb4e55bda3ad594b7fe6f4bb01410d5d5568cff",
   "tracks3d.npz": "0301b04e3861aa0b9a89e8cee8c5add79b127e31daed1920a99e14d77f94675a",
   "videos/view0.mp4": "2360d1b1fd675db96638642e001553f0584eb5b425d308c15bd4d24cda883e9a",
   "videos/view1.mp4": "045576e1d7d7f368312f53d3df1849d8d373706de4a4d4618dcb117afdd6c488",
diff --git a/Simul-Robot/c001_episode_004660/source/cameras.json b/Simul-Robot/c001_episode_004660/source/cameras.json
index 9485c80b33e382edb1d9ce062aee8e85ac7a9091..b88c2c7e18efd551c532abc62d59cc4ae65fa421 100644
--- a/Simul-Robot/c001_episode_004660/source/cameras.json
+++ b/Simul-Robot/c001_episode_004660/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_004660/source/meta.json b/Simul-Robot/c001_episode_004660/source/meta.json
index cc1b053e6df04aef84d7c884fd50bc5fd1251050..61963e689f22f5b79dbaed0e77a6305e61a1cb7e 100644
--- a/Simul-Robot/c001_episode_004660/source/meta.json
+++ b/Simul-Robot/c001_episode_004660/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 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']: [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: 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_004660/source/visibility.json b/Simul-Robot/c001_episode_004660/source/visibility.json
index 6fe7e0fce4697505bc9623d9d28fb0b78f51c404..94194664d52ccb5ce7ea78c0dd7ea13d68cc4bfb 100644
--- a/Simul-Robot/c001_episode_004660/source/visibility.json
+++ b/Simul-Robot/c001_episode_004660/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": {
     "potted_meat_can": [
@@ -5113,9 +5113,9 @@
       "target": "potted_meat_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 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']: [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: 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,
@@ -8503,5 +8503,5 @@
       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_004930/sample.json b/Simul-Robot/c001_episode_004930/sample.json
index bf2fbaf81daa28e116edfe020820d360510a8c38..c29af7a14a176f1b76126b862596c1a539f583d0 100644
--- a/Simul-Robot/c001_episode_004930/sample.json
+++ b/Simul-Robot/c001_episode_004930/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "7a746ced37dbc4dc821d5aba1eb963ff6d42575d8d8657b30262b73d6514fa03",
   "labels/objects.json": "72486c69f152fb3119fa1519e7dc54ac564932f4bbbf38f2eb03a6360e03030d",
   "labels/visibility_states.json": "a7be20483b17555693e731273c74855dc1534398d5becf8197980b539a2c2c8d",
-  "source/cameras.json": "7704cc6bff7de2c781434da1273ccb95c6b97e38c6cfa84ca217fb5be05f689c",
+  "source/cameras.json": "f84027216fd6591628c7b034127659d1e54366dded557bb6a264eea4689f834e",
   "source/caption.txt": "393a85e9179c92b8ca366943d83e1e5dca49e56b0a08cfbefb1c2d0c18d263e8",
-  "source/meta.json": "3bfdd4a53b158143b347bf69175b503feeff1ae55632dbb7ef45068276d23bd8",
-  "source/points_target.npz": "8254f629a23af85b591e16728e813285434a3a3fbf9e432d184a184f642a5adb",
-  "source/visibility.json": "e7d6fd292d2b12feffb502b97ca99d96dc300e390920678d3524ea8f67750af2",
+  "source/meta.json": "96eb2cf2994098db796155c3094dabeb32ac557eb52244348de6d001c44bff2a",
+  "source/points_target.npz": "f6ebf429e082c2ed49d536e90fc1d20065f5792f3f56cacb7c5f2a49a985cf02",
+  "source/visibility.json": "98ab8ac6268071d279037e4f17ef8eee999f660781802c98290dc35210d269e9",
   "tracks3d.npz": "58c07d02254fbd5f2601987e9f5c52563199c37a04c2987382926c31740f8b62",
   "videos/view0.mp4": "69198e27ca8103d558bc49f9b669fc4d8e57491c157a8355f8938433a3294d4d",
   "videos/view1.mp4": "3d53f12348678d66c7db87f7f550caa6fc57e0b5c76b9e76c6dc8cdfba9207cc",
diff --git a/Simul-Robot/c001_episode_004930/source/cameras.json b/Simul-Robot/c001_episode_004930/source/cameras.json
index f31aa7fa000bceab283a4d6108733df8c685074a..7419232b22753ef5398a462b119a3db80d09a640 100644
--- a/Simul-Robot/c001_episode_004930/source/cameras.json
+++ b/Simul-Robot/c001_episode_004930/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_004930/source/meta.json b/Simul-Robot/c001_episode_004930/source/meta.json
index 7ce51f41c95853a6d85c8ece1e755ce1b49eddbe..7df5b5f6b3ab14613e1b402ea0688c8e5e5c75c7 100644
--- a/Simul-Robot/c001_episode_004930/source/meta.json
+++ b/Simul-Robot/c001_episode_004930/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_004930/source/visibility.json b/Simul-Robot/c001_episode_004930/source/visibility.json
index 239e793f28921d1e74487f5649423b29ddbb48d4..e4fb9d10aec4c681abda52ad7b798ea3f3da2243 100644
--- a/Simul-Robot/c001_episode_004930/source/visibility.json
+++ b/Simul-Robot/c001_episode_004930/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": [
@@ -5590,9 +5590,9 @@
       "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,
@@ -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_005090/sample.json b/Simul-Robot/c001_episode_005090/sample.json
index a1ad0ef44a91f03172b2d24a03eb09bc9b9a070e..074411d4515eb3ecfda72cfc40962e460c8edc07 100644
--- a/Simul-Robot/c001_episode_005090/sample.json
+++ b/Simul-Robot/c001_episode_005090/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "e2b65231ed1d7724b75162a8e5656003ed4f552beb2c0ef1c24cdb39bb023497",
   "labels/objects.json": "daea187ebf8239bcdf5008b93a22848dfe9e7234aa1c1b1f392b6df2aec2fd7e",
   "labels/visibility_states.json": "d5e60f9e404c43fb6330bc48e5cc0c22ea0386401ef94326ec647f6bbcc05a7a",
-  "source/cameras.json": "bc9f19cf2d4cd3637cade6a760b10ac924fb4c69945f744cb4c198aec62be31c",
+  "source/cameras.json": "f2040a274d37f66aeee0b486c8162f93d0eaf54cb275ea1b67da855628ca8b37",
   "source/caption.txt": "a17af15cc9052c202c217b55079463552d68f87ff5194b0915e590f6c08b5294",
-  "source/meta.json": "8b750d48884da2550b43a5e1e1ed93817a04e46c1409cd1b9efeaf6ca82607b9",
-  "source/points_target.npz": "7b70f3143c7c2666109620292c29745d33afffad8e4a8b2076510ed74d0e81ff",
-  "source/visibility.json": "2938070cada5ea0bb85f506e845ed13cb3513c57fe3efa2da81695198df845c1",
+  "source/meta.json": "61a266df5f980977b827f3189ef1daa9cfa8f1c808d2e1540af2fa53a2078c30",
+  "source/points_target.npz": "cb05ef55ff18a345b50304b83ea7eebd636e8c76c65e64f68d4850132f129748",
+  "source/visibility.json": "c6501bdc08316bbca57088003294c54c59e77c573f7c0c61d71d00414edfb65d",
   "tracks3d.npz": "4da6b570c0554fa50e9e819c8a2e3d4a6f3cd77770ec11a117a9a320e5fec18a",
   "videos/view0.mp4": "a9d910948d418bb51f3b33964f7232b9e392c4da2b2abc6a3345691d8fb266cf",
   "videos/view1.mp4": "d40a4977afa7c7aa7c0ee9edabd8f15882daab716c77c3dd300ae059d50dc2dc",
diff --git a/Simul-Robot/c001_episode_005090/source/cameras.json b/Simul-Robot/c001_episode_005090/source/cameras.json
index 82b7a21b567c57f1093b5a1a99e5cc50beb22b18..1c9de2e1df6c502031831bfee6afa965e76637aa 100644
--- a/Simul-Robot/c001_episode_005090/source/cameras.json
+++ b/Simul-Robot/c001_episode_005090/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_005090/source/meta.json b/Simul-Robot/c001_episode_005090/source/meta.json
index bf5f5bf6bb4da997ffe6cb06ab2aa5c8c227248c..9bc4765593078c850f9a8d894d2bb2cdcf4f9662 100644
--- a/Simul-Robot/c001_episode_005090/source/meta.json
+++ b/Simul-Robot/c001_episode_005090/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [15, 1, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.17, 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']: [15, 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']: [0.17, 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_005090/source/visibility.json b/Simul-Robot/c001_episode_005090/source/visibility.json
index 2e3ec9c4fd1f78363d898fe3c272cee4fdd10ac4..6e1df02c045eaad559104a99f499c03790c174de 100644
--- a/Simul-Robot/c001_episode_005090/source/visibility.json
+++ b/Simul-Robot/c001_episode_005090/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": [
@@ -13477,9 +13477,9 @@
       "target": "painted_tin_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [15, 1, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.17, 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']: [15, 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']: [0.17, 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,
@@ -13489,12 +13489,12 @@
     },
     "relay": null,
     "hard_target": "painted_tin_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 14:01:31",
       "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": {
@@ -23559,5 +23559,5 @@
       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_005210/sample.json b/Simul-Robot/c001_episode_005210/sample.json
index 85e2ea6efa0bbd7ba0c2a628f93fda7d1e5e652c..97f6eafc74d6638f52bc6dec1af36e8469f3a09d 100644
--- a/Simul-Robot/c001_episode_005210/sample.json
+++ b/Simul-Robot/c001_episode_005210/sample.json
@@ -218,11 +218,11 @@
   "depth/depth.zip": "8d17dabda3f87dce4d0cf9b600672088339dfc4f7d6e0885b675c06056d09ee6",
   "labels/objects.json": "c27eb5a9f45d58cb4e3aa8788ae9d6fa1f2cee7749c31d1b66a845dfc22c34ed",
   "labels/visibility_states.json": "8a1a7183d70d5ba33a25844d0deab14168d08ba5ac553fcb58799a20d24f8055",
-  "source/cameras.json": "850a81a63705732d52e95513ba5594baeb0143d42934a78a715eb7e24b3df305",
+  "source/cameras.json": "d1fdfd217b4d4b532e8cc10f5ee711cb6244ea8bd7347fbdde89c9180f016b97",
   "source/caption.txt": "30f9895e183605217881521cc74627df39ff464a07e1edc54c75012fe67aac26",
-  "source/meta.json": "60f94c87ce148ac31cd249242ef1929a2471ea1499da2699ffae0c4494ef7be5",
-  "source/points_target.npz": "f42cbcad81f3f05a678ff0006312c220a645cf90768ec4b861cbdb2b34bbfb7d",
-  "source/visibility.json": "ec12d087a6fca9c19f21e0b03f3157fcb95fa508864905185b2edf5de2c4f62a",
+  "source/meta.json": "84501eb1c905db6a52fcfc0e4a1c90781cd60148c5e3fced4ebbd96bd74de581",
+  "source/points_target.npz": "c361eb1428b4fac86928c8b4c8a5c296419059f216068008499a06839af44bde",
+  "source/visibility.json": "f5127eb6a89f0e99ac67b332da0341144047b00dce25fcb6ca7324d5972e8dfc",
   "tracks3d.npz": "86db770d3c9049ccb6918294fca1332ee122e02980a3fb6889ab332ce53e713f",
   "videos/view0.mp4": "7664e8796ea0e0593dd7e3404a2ed35fe6acf066d0c4dab88cee7576b0a70ab5",
   "videos/view1.mp4": "d4ea038d08ea76e8a2c43d46c45844ec952c13194401f41945683520fd25635f",
diff --git a/Simul-Robot/c001_episode_005210/source/cameras.json b/Simul-Robot/c001_episode_005210/source/cameras.json
index 1618bc53d805ffef5533cd710ba37852ab2c199e..1ff25a4c77641d5f02a0cf94e4a5e7a5b9cf1e3c 100644
--- a/Simul-Robot/c001_episode_005210/source/cameras.json
+++ b/Simul-Robot/c001_episode_005210/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_005210/source/meta.json b/Simul-Robot/c001_episode_005210/source/meta.json
index 9e0ca175b3e2e5f4c80e982c4cd5f39b5b4d7ebe..695bad45413c91f90bb38aea76ab3dd59cce745b 100644
--- a/Simul-Robot/c001_episode_005210/source/meta.json
+++ b/Simul-Robot/c001_episode_005210/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 2]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, 2.19, 1.24] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 2] 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, 2.19, 1.24] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_005210/source/visibility.json b/Simul-Robot/c001_episode_005210/source/visibility.json
index e2f192dc337e0c01ac4b4769b6ba63245c9ad001..284e19cc1e882ab9702839742ab3575110f7dfc6 100644
--- a/Simul-Robot/c001_episode_005210/source/visibility.json
+++ b/Simul-Robot/c001_episode_005210/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": [
@@ -15956,8 +15956,8 @@
       "target": "small_yellow_cube"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 2]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, 2.19, 1.24] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 2] 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, 2.19, 1.24] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -15967,12 +15967,12 @@
     },
     "relay": null,
     "hard_target": "small_yellow_cube",
-    "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 14:01:36",
       "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": {
@@ -27897,5 +27897,5 @@
       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_005460/sample.json b/Simul-Robot/c001_episode_005460/sample.json
index 7689e00135cd2d9a2693034c50d2c2acadfb2cda..aa103445f2c7f76741ba47bd96a4974c3f6e5c34 100644
--- a/Simul-Robot/c001_episode_005460/sample.json
+++ b/Simul-Robot/c001_episode_005460/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "65f9998286fe009942bd6219dba9283fdfa3a75d0439c556ee191b546a73391d",
   "labels/objects.json": "9ebe3f26d5b02e5bbd3e979986e73d2957891381adf34295849ca6135004e36a",
   "labels/visibility_states.json": "5dc151bfb1a3143f292fae38f28c5a2d2609645a66f8bc058873e7edd4cdd58f",
-  "source/cameras.json": "21f73bf14f48fbd841737b11ee000bb73975a45019d64b4b52b3b260a8d9019b",
+  "source/cameras.json": "e88aefeb6465cab10ac91e4e63f4df1becae1b033d63c84d6345a91db49c8e36",
   "source/caption.txt": "550ee90f388c1469fb6f33b92eea93fe2904fdcd47488fc45618fd5a80145e32",
-  "source/meta.json": "1ac4cec9efecf3cdcf9405616896fc749adc2f4c3da718408b40df073ef8d712",
-  "source/points_target.npz": "a54f9ed13760e182a4e13110acfd7d397dd1fbf98ee1fef1621565a1e87a70e6",
-  "source/visibility.json": "4c83751a82c812ab6865fbda9fd04ba4da7555d50134f974596b0a74ddb2bdde",
+  "source/meta.json": "538523aed6d7a98414b65cf43034ac246e35f30a7c9324c46ff3ae3347136438",
+  "source/points_target.npz": "5263ea34e43208483eadf50615ed6c443633ab441bb007b1570a2536663db31f",
+  "source/visibility.json": "82297ae90011a1e542f1b129fedf4a13d6030c9ffe883d81d47aa2255275a1f5",
   "tracks3d.npz": "455a2ec2f28d299aed65f7f98d34cba9fb239df38b5a38872a915ed3a57c0d08",
   "videos/view0.mp4": "a89f83b0bb4d3ec05d746b165d56edb0742b00fd38775b86ba210cecb65e3115",
   "videos/view1.mp4": "7486becea642e2fbf62835aa455fcab7273cc715510eeed217e505486c62569f",
diff --git a/Simul-Robot/c001_episode_005460/source/cameras.json b/Simul-Robot/c001_episode_005460/source/cameras.json
index 8f156091ccc06b7ef7632f6732c9cb61ba0e2a83..b99596198163f7d4e006a00264ad7304d51a2b05 100644
--- a/Simul-Robot/c001_episode_005460/source/cameras.json
+++ b/Simul-Robot/c001_episode_005460/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_005460/source/meta.json b/Simul-Robot/c001_episode_005460/source/meta.json
index 1ad2b53ea90b6192f20a0e3ae6a5191e8207d223..61e7d5d03f38cbe50be5f7cd26856f41f105d16e 100644
--- a/Simul-Robot/c001_episode_005460/source/meta.json
+++ b/Simul-Robot/c001_episode_005460/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 [0.0, 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']: [0.0, 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_005460/source/visibility.json b/Simul-Robot/c001_episode_005460/source/visibility.json
index ad851e9a640009ba4462066554fd943f04407c67..a808378a0ba2672f7116b3839db242f37e61c02a 100644
--- a/Simul-Robot/c001_episode_005460/source/visibility.json
+++ b/Simul-Robot/c001_episode_005460/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": [
@@ -15154,9 +15154,9 @@
       "target": "blue_seasoning_shaker"
     },
     "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 [0.0, 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']: [0.0, 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,
@@ -15166,12 +15166,12 @@
     },
     "relay": null,
     "hard_target": "blue_seasoning_shaker",
-    "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 14:01:56",
       "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": {
@@ -26496,5 +26496,5 @@
       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_005600/sample.json b/Simul-Robot/c001_episode_005600/sample.json
index e323e5ce48e84cc4f6dd072c36489a420b54d163..778da4eb8eab6256cb4b80bb889521e082c28a49 100644
--- a/Simul-Robot/c001_episode_005600/sample.json
+++ b/Simul-Robot/c001_episode_005600/sample.json
@@ -218,11 +218,11 @@
   "depth/depth.zip": "ed077e40b3607d3de827a49af9195a26925e619762ca22c493ce6e0e772f4214",
   "labels/objects.json": "988217719a5ddd4f19783795f1b02ae5292d0c717ede9d11ab44afc12cb579a2",
   "labels/visibility_states.json": "9eeaf67645d49c2e5368479d46c7144018453742636b5c749264022986f6f705",
-  "source/cameras.json": "4370628758f52aedd40904b5b93dad63b801fe10a9d7d2f78c4c42c505564695",
+  "source/cameras.json": "0188ba1e362bc1664a0f39346d58930e3c4170b2e67e9a8922d92177d6734afe",
   "source/caption.txt": "ee5bafcd53397b4f72bf076b015015064647685eb6c14c2041130de799ac7ca4",
-  "source/meta.json": "f08db4cad7fa690a1cc18602666b82743334e24f0492443374daf81cbeadf3cf",
-  "source/points_target.npz": "d430025b4aab2aede3da4cd8b59dd8c12ee21d3482c94611ff18c908e90243d0",
-  "source/visibility.json": "a0e5a1937a31368703a34b874056dae44bd6a73871bc19f7eb10bfc5eaacace5",
+  "source/meta.json": "4754bafee347f11a6471d1432bec94dd0e4ab89f7e3b0e211da5f21d31480cc4",
+  "source/points_target.npz": "22163cb39d281c8f94f026411cd97cb3a2459cf7959a2a33f98773f460fa9572",
+  "source/visibility.json": "ae901721ec8d3b2fbf3f68a8312cc071d5836d66986a6c43f2659b734763354d",
   "tracks3d.npz": "7253daf5d4f38882562d203a8f2f2f565fee200b084c635fb7b2a11d5100043a",
   "videos/view0.mp4": "0a1368e20a172fa6893d43d043e3f7696477a7c1c482c36c1f56c2acceba72b0",
   "videos/view1.mp4": "88ad9a20f5f38fcd1409e243893bd8b697806081fd3d028cf71a32e15ae65c6d",
diff --git a/Simul-Robot/c001_episode_005600/source/cameras.json b/Simul-Robot/c001_episode_005600/source/cameras.json
index a097f5c9c863f9148d03809bc389de8668b3c7c0..43606876cc01933af39fc2ae928755578400d3fd 100644
--- a/Simul-Robot/c001_episode_005600/source/cameras.json
+++ b/Simul-Robot/c001_episode_005600/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_005600/source/meta.json b/Simul-Robot/c001_episode_005600/source/meta.json
index 42f53b56b6a00dd4544b7ea07805fd2b40e85629..ad0efb5e5734d94b1019408c5be4698e4d47e73e 100644
--- a/Simul-Robot/c001_episode_005600/source/meta.json
+++ b/Simul-Robot/c001_episode_005600/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_005600/source/visibility.json b/Simul-Robot/c001_episode_005600/source/visibility.json
index 1946edc72c736349bd0837b0c6ee1d7f62c5aa3b..059ec5c6e0f3385fd2e487018a629a729a8b4e40 100644
--- a/Simul-Robot/c001_episode_005600/source/visibility.json
+++ b/Simul-Robot/c001_episode_005600/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": [
@@ -15313,9 +15313,9 @@
       "target": "blue_seasoning_shaker"
     },
     "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,
@@ -26767,5 +26767,5 @@
       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_005620/sample.json b/Simul-Robot/c001_episode_005620/sample.json
index f62df15339f4285cf5ed7d20d1a528d5e1662d06..c1eee55ca31c6b940f9dda7986ae5f2b80e3f1f0 100644
--- a/Simul-Robot/c001_episode_005620/sample.json
+++ b/Simul-Robot/c001_episode_005620/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "7462abd61ab60dbcb9c86690f9f03253b106bd461125e4e709339863fe9c6242",
   "labels/objects.json": "5ea0e23ee7d688ee06f0b6d0d21fcdcf6c0b03206a2c1682981bd3e5dc91408e",
   "labels/visibility_states.json": "53535844ec222bb862be805f873a7ba83ca24faf37c375c1a99cea1ca8810fc3",
-  "source/cameras.json": "3350037d570b2d52503953b0eb3f4ec41511754ad7935360d8efaa2e81f601b7",
+  "source/cameras.json": "815864b3f9c5bb413ce75114d3148c884bfee06cd68a8714b0fb0e01cd69dff3",
   "source/caption.txt": "664a511ec5c97567b2a13762aa4d2469367889ba63a0e0a1f1779526f2831b96",
-  "source/meta.json": "7698c12f356c44d7933b5c9ed9fdc33b9732b742340786735773b812e4a891f4",
-  "source/points_target.npz": "f8d5fd5f90ee63cad984b1912c0c8d8d34e5c312bad65ecde51acb11b28f5932",
-  "source/visibility.json": "d6e3a3ca3e4ff3db3e7d015bc2de0b00c92f97df4d80b8bed973fb311d248d35",
+  "source/meta.json": "5288658c43e268bf0be8f9bfe27a3f5ba9a54f0aeb219624f187f6173dea3bbf",
+  "source/points_target.npz": "dc068b3f22a54b5039df6cdbcb9d7decc68175632d0d548e13ec65ce3052489e",
+  "source/visibility.json": "1d3b170772b7e803bb6183c1abe910361988d6008882ef46ec4b87d6ac030ba9",
   "tracks3d.npz": "5b0d21a07aadb723aa3dfbbaa6e343c512d1ff00405f31eda208e2b020c48106",
   "videos/view0.mp4": "9c2d4cb8037e10202085cc6228d47a8df12a29d09320566b203d44d9cb36e25b",
   "videos/view1.mp4": "5cb1507adf823eb344d2205ac9eb48ee51980e114b37f9f161031774ce3db022",
diff --git a/Simul-Robot/c001_episode_005620/source/cameras.json b/Simul-Robot/c001_episode_005620/source/cameras.json
index ecd0bb161120100fcb96083781494e8fa02fb2c7..788f141e6f68e537f7706898e1b2a134e3feb1e4 100644
--- a/Simul-Robot/c001_episode_005620/source/cameras.json
+++ b/Simul-Robot/c001_episode_005620/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_005620/source/meta.json b/Simul-Robot/c001_episode_005620/source/meta.json
index 91e07a99dad530018225bc1c97131aabcf1cb93c..ed6f8d45fdfc05b1120e8f14f859c2e607afa1b5 100644
--- a/Simul-Robot/c001_episode_005620/source/meta.json
+++ b/Simul-Robot/c001_episode_005620/source/meta.json
@@ -41,9 +41,9 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [5, 4, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, 4.75, 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']: [5, 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', 'cam01', 'cam02']: [0.01, 4.75, 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/c001_episode_005620/source/visibility.json b/Simul-Robot/c001_episode_005620/source/visibility.json
index 0adc21bc76c53b482c9a0cb3ed0aa9a1e76a0f6f..2c1b129e9d5d6ab5f5deebc1ad6d46b8cc1b9316 100644
--- a/Simul-Robot/c001_episode_005620/source/visibility.json
+++ b/Simul-Robot/c001_episode_005620/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": [
@@ -16477,9 +16477,9 @@
       "target": "glazed_squirrel_figurine"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [5, 4, 0]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.01, 4.75, 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']: [5, 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', 'cam01', 'cam02']: [0.01, 4.75, 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,
@@ -28801,5 +28801,5 @@
       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_005740/sample.json b/Simul-Robot/c001_episode_005740/sample.json
index 2598c0fe4185d856256940363cd6d5d5de8ab6d9..1457c31e7baca2d45a4d274c2da6d98d8f770bac 100644
--- a/Simul-Robot/c001_episode_005740/sample.json
+++ b/Simul-Robot/c001_episode_005740/sample.json
@@ -218,11 +218,11 @@
   "depth/depth.zip": "80e649981f1887f19b2dadc6d08ca83a4f44e5f6487b70195d4f4c1a1bc68fee",
   "labels/objects.json": "88b8be806fd1e48923e2ef2c51eca0383c6fb1fb169d477d77e953c522e4e6cf",
   "labels/visibility_states.json": "bbbbefe419ec6dae00c4ca1aad9a847ee2eb2347e587f7d66cbb33f4c7d80c4c",
-  "source/cameras.json": "e38e3d92399304bcd4a54cdc736ccb1ab7776d392c4b09239ea0e7538d0efbef",
+  "source/cameras.json": "11212cf71cf498e27f2c9d5549fd3f210b877823fc6ce60b26902825632bd238",
   "source/caption.txt": "8cb9f8bd6903dfac44ca364f4203a7e32993c99b88918f90d26d53469785fb4f",
-  "source/meta.json": "976cdf155dfac27dc425ff8ee77a8683aae0ce42f50237cf04118b13f4bac13b",
-  "source/points_target.npz": "2d27bff94f8864166855a69c5374632341f22ee5fd89632eba9db333b37dbfdc",
-  "source/visibility.json": "2866848395b6679da175d5a0133be212256bc6e759c9ee4a23c6d5c1396ccdb2",
+  "source/meta.json": "4cd42ec1f7f9312a37efd80436bcdf2bc8666843c7cd5ce4572db49a2e200556",
+  "source/points_target.npz": "ee82913b2ea65c44cc41526ec9327aa831caa6c1c708e3879847b2298b7fd75d",
+  "source/visibility.json": "6385127247836b32bc3c52ab97f1299fc05821099c1ea5a767f34a3820478c58",
   "tracks3d.npz": "f3dfa8cc1724bc9645b708e90e2b5dddcce26008e5f8a535109563fecde7e43a",
   "videos/view0.mp4": "774ff751325ce372049e63614154312d21d00baddb7b00fd1c23f3e31ec4555a",
   "videos/view1.mp4": "8d242fb7fe2ab42f83393ce02ebe6c2a6186e2ecb3454550fdfec59a8fb53670",
diff --git a/Simul-Robot/c001_episode_005740/source/cameras.json b/Simul-Robot/c001_episode_005740/source/cameras.json
index 58d67a48b35f4974a839bd7024b92f01fb29a0d4..0a1149a8febb2d94d15e49dffe8f5f557de284c6 100644
--- a/Simul-Robot/c001_episode_005740/source/cameras.json
+++ b/Simul-Robot/c001_episode_005740/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_005740/source/meta.json b/Simul-Robot/c001_episode_005740/source/meta.json
index 537d5214958cd351e7d5fc9a1d77cb48498e97c5..cf9279f7de0d06909ada8f26cf7930ff8db07469 100644
--- a/Simul-Robot/c001_episode_005740/source/meta.json
+++ b/Simul-Robot/c001_episode_005740/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 1,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [3, 2, 13]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.42, 0.12, 0.05] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [3, 2, 13] 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']: [0.42, 0.12, 0.05] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_005740/source/visibility.json b/Simul-Robot/c001_episode_005740/source/visibility.json
index 29934a5ecadcedb7c36509643a2d104f4393fd41..3d8f73b289f0f0084069b27d63e90850b54d2310 100644
--- a/Simul-Robot/c001_episode_005740/source/visibility.json
+++ b/Simul-Robot/c001_episode_005740/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": [
@@ -14199,8 +14199,8 @@
       "target": "painted_tin_can"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01', 'cam02'] 의 최장 비가시 [3, 2, 13]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [0.42, 0.12, 0.05] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01', 'cam02']: [3, 2, 13] 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']: [0.42, 0.12, 0.05] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -24812,5 +24812,5 @@
       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_006740/source/cameras.json b/Simul-Robot/c001_episode_006740/source/cameras.json
index ca7109e934ddeb98580c9cbbb0d2e9b31b6ca662..2a074e72f04ad4cce5284a58e509ba78394c7873 100644
--- a/Simul-Robot/c001_episode_006740/source/cameras.json
+++ b/Simul-Robot/c001_episode_006740/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_007280/sample.json b/Simul-Robot/c001_episode_007280/sample.json
index 0a6de41fc79a27b68bf88866895eeadc54724068..f50e777a9e533436f725732ebe910466295a1bac 100644
--- a/Simul-Robot/c001_episode_007280/sample.json
+++ b/Simul-Robot/c001_episode_007280/sample.json
@@ -219,11 +219,11 @@
   "depth/depth.zip": "8b130c62517194c338509e6e37f21fd1018411a90ac782f37966ce178ee01a74",
   "labels/objects.json": "1791ffcdbefd671d9805bc33b721bb350d450dcac10b5caa4dbfb09bab46d4e7",
   "labels/visibility_states.json": "cc51d180a40365af5947430fc0d55eb2537eaff50967c832bbb5b21fe86238a6",
-  "source/cameras.json": "dba09d950f80d746f4a16ab210e19ed73e89c0e5eea5e472858ce29957dfc827",
+  "source/cameras.json": "f716a5dda58a00925166171ddca81b8289b092d6480ead9188e0f6a09a08cc4d",
   "source/caption.txt": "e95b636bb625018fe14b75adc67492688e27cb7e3b1409a9a8e929e710cbb516",
-  "source/meta.json": "9d9d8204b1596474590aa38534995b02766740d8ed635d7627058c186833d7ff",
-  "source/points_target.npz": "66e5452181e9e33ab59285730bc81cabb500e7b21376d2843bf632d72343ca44",
-  "source/visibility.json": "9a752c5fa9c1d48131206fb63b08e0a2b891aed84446af08770b1ca50439e16e",
+  "source/meta.json": "7d67de7b28ca6e7b3b02123cda1dbff8180dd0891a9f9b75bf274fc51080b0b3",
+  "source/points_target.npz": "4b14c3b8b247e0d0208991f858d480f5a552b5c6707f4f977a72cd1a3cc8e650",
+  "source/visibility.json": "c301ef120a57b7789f15d49f9a5f4154ee8bcd8714d3cf59e1bb2a213ef1c60c",
   "tracks3d.npz": "e4ab5b14e13d14244601be25848df9891799aa82ecf978972d14b3c38a93b6a2",
   "videos/view0.mp4": "cb6f3ef16ec050d06c7711f6db24d5e1f39f821dd5af392782f3651739f7daec",
   "videos/view1.mp4": "de4ad7e06133bb730d6c4b2ec756893879bb53bb9ee97bef27a16c333b107b02",
diff --git a/Simul-Robot/c001_episode_007280/source/cameras.json b/Simul-Robot/c001_episode_007280/source/cameras.json
index b35f98ba3ce231484e4e9c2c2abf1b125d8a9be4..81d9c345e38a1d4ffefd3f464f6ca4aa49043189 100644
--- a/Simul-Robot/c001_episode_007280/source/cameras.json
+++ b/Simul-Robot/c001_episode_007280/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_007280/source/meta.json b/Simul-Robot/c001_episode_007280/source/meta.json
index bab730e4cf4be0c2a7b3389d8a5c82fa871fcda9..25a45c4ddd872b4d63232585fa9c326cf53779c7 100644
--- a/Simul-Robot/c001_episode_007280/source/meta.json
+++ b/Simul-Robot/c001_episode_007280/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 [1.42, 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']: [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']: [1.42, 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_007280/source/visibility.json b/Simul-Robot/c001_episode_007280/source/visibility.json
index cc5b01a16030cddf307dffbd41f53a3ffd223c74..e84a627349b78b2db9d6038ad185e23fa68d8825 100644
--- a/Simul-Robot/c001_episode_007280/source/visibility.json
+++ b/Simul-Robot/c001_episode_007280/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": [
@@ -15994,9 +15994,9 @@
       "target": "yellow_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 [1.42, 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']: [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']: [1.42, 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,
@@ -27958,5 +27958,5 @@
       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_007300/sample.json b/Simul-Robot/c001_episode_007300/sample.json
index 6d39c5de942b607b27d2159e334dd413b045ae13..880ce0541035c76b0f8bb9067b2b7fee251d447a 100644
--- a/Simul-Robot/c001_episode_007300/sample.json
+++ b/Simul-Robot/c001_episode_007300/sample.json
@@ -218,11 +218,11 @@
   "depth/depth.zip": "c421f3c4dfecdb4db0676971e3b68b9421fc5569d82c16e609ce32fa636c0f9a",
   "labels/objects.json": "e5e1397d7c0e27e586e984874987ba2ae5f9234bed44865b43f3b4c57b63fbdc",
   "labels/visibility_states.json": "1d17593ce602228cb4546381d70cc9ee2ffccb65299b2c9d0f7708330df1bf4d",
-  "source/cameras.json": "da1ea8aaf9acc113527ede4d2b61bf4f248af42f58c0aca24707c0d3387445a9",
+  "source/cameras.json": "d5dc5d1a172acf9806754cad7dd5a2613d980d67aa61e1e537f235bebdd39473",
   "source/caption.txt": "e1da4147f7b6891d10278dbda278bcecc17daf4419995bdb1751946c8cffe577",
-  "source/meta.json": "2a4cb89860fa02790aa50cfce992f47c1dac888e538b809551b6739106ddc1e0",
-  "source/points_target.npz": "45043fd25b7536562725d7ec40dee9f2efcb3cf19d949e85edc3382e5a90afb9",
-  "source/visibility.json": "d4c2c4fa5f9942cd17ecddfd5e68b2fd2cb3acf24999ead7b90f38c798746ab3",
+  "source/meta.json": "5a1accbf69b41c1b11fc7c11ace9ee581c5a002cba1cb607400970542ebbfc4d",
+  "source/points_target.npz": "5e58b6e00f59f5ef83375c22109aae09fc180119e500bf6b52dbfe44e13c230b",
+  "source/visibility.json": "833b38e7c818d62f2ee7f35d290830741351dbf34fc371c228b52cf5ec76d782",
   "tracks3d.npz": "49d6b62176f0a325561d9d84fc3f2c8a38704ed78ebe96abde04a532ae0b1d10",
   "videos/view0.mp4": "845647d1b952a760b45d495349cd16b461a6b7425d3ce612cb6167049272bc20",
   "videos/view1.mp4": "89c4ca8f88dabf28bccfbc85854c0a2bd56944a69a83d3409034586dc4c0402f",
diff --git a/Simul-Robot/c001_episode_007300/source/cameras.json b/Simul-Robot/c001_episode_007300/source/cameras.json
index 6707a2e0c7f9ad4643caf6069845fba8db5f7a4d..d44a26d6e96ac239d7de2ad23c8644c16cab7034 100644
--- a/Simul-Robot/c001_episode_007300/source/cameras.json
+++ b/Simul-Robot/c001_episode_007300/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_007300/source/meta.json b/Simul-Robot/c001_episode_007300/source/meta.json
index e3378f792f61006665abfa3c380cae5b4046249b..86102b42ace2e9cef554f27ce4be65aa12e4badc 100644
--- a/Simul-Robot/c001_episode_007300/source/meta.json
+++ b/Simul-Robot/c001_episode_007300/source/meta.json
@@ -41,8 +41,8 @@
   "hard": {
     "hard_k": 2,
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 3]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, 0.0, 0.78] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 3] 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, 0.0, 0.78] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_metrics": {
       "num_cams": 3,
diff --git a/Simul-Robot/c001_episode_007300/source/visibility.json b/Simul-Robot/c001_episode_007300/source/visibility.json
index 04d5c7f0098923992050abf2fee329c098cdc4ab..85a2618019cffe7869e9979de84d7886ad7eea68 100644
--- a/Simul-Robot/c001_episode_007300/source/visibility.json
+++ b/Simul-Robot/c001_episode_007300/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": {
     "brown_ceramic_bear": [
@@ -13156,8 +13156,8 @@
       "target": "glazed_squirrel_figurine"
     },
     "hard_reasons": [
-      "①캠 ['cam00', 'cam01'] 의 최장 비가시 [0, 3]f < 30f (그 캠 하나로 전 구간을 볼 수 있다)",
-      "③캠 ['cam00', 'cam01', 'cam02'] 의 비가시 구간 타겟 변위 [None, 0.0, 0.78] < 5.0cm (못 봤지만 잃은 것도 없다)"
+      "① longest invisible run per camera ['cam00', 'cam01']: [0, 3] 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, 0.0, 0.78] < 5.0 cm (not seen, but nothing was missed)"
     ],
     "hard_thresholds": {
       "hard_blind_min_frames": 30,
@@ -22989,5 +22989,5 @@
       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_007650/sample.json b/Simul-Robot/c001_episode_007650/sample.json
index c523c7a7dae32340847f095f71f96e043efbfc25..dd356ecb831b93d0ddd056200849c03653a8c804 100644
--- a/Simul-Robot/c001_episode_007650/sample.json
+++ b/Simul-Robot/c001_episode_007650/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "9be9160c3eb49dce48cc705006f9b51e599c271bbc79058e582e937556e0587b",
   "labels/objects.json": "7f82e49351ca17653d39992c313c93bb3c1678acfa7e6f56e25e3a2f270cac35",
   "labels/visibility_states.json": "221dcf4f8fdb7bb6a8a8cd0be07130bdf75baa8fac91cdf85d0225b8bc953a8a",
-  "source/cameras.json": "f23ee799ef23ce57e0a0e2cd932772cf36e487f4ac439f221915e5538edd9c0e",
+  "source/cameras.json": "b6a64c48f825f6aa1dcd5bde3581d7492fddf3888128d820a49dd2ebcad0738e",
   "source/caption.txt": "a6f9378b78d613ef92a79238c9075c6df944e190a0691cf07c3288a6ac46dd0d",
-  "source/meta.json": "a54669c36f078bea614ce017c029a12c73e6cc158b20f9098d91f45636dcf402",
-  "source/points_target.npz": "3e4cd70963c86a64f14791f08ccf3fec3907f035b6f3342728ae375ca84a6106",
-  "source/visibility.json": "521cdeed3dcabc666ba72c892d575a6ba0e7f613e0db52aa9c756c6b0c5d693e",
+  "source/meta.json": "078317c344826442a55f66697231071dac823c5c9f09f30bd88de877a99bbb47",
+  "source/points_target.npz": "9e9f4a433563982960681cd6f79461dceb40ad721f403f7c844135697586478c",
+  "source/visibility.json": "0e9577d57ae922a8d42c4b1e70a816adb5d165e6b70861207991dac4c0a3da68",
   "tracks3d.npz": "709fc9243537a8f41655ff022ac40ed084ebcc781197d4b431b4ba641e563e74",
   "videos/view0.mp4": "59331f9e771241fcb6b65af0e321e9ea71bb184e17fe5ed755ac48ece55d4da8",
   "videos/view1.mp4": "27270d52cf7492fa0101d4a9cad9d3b805450a337c005ae4de4b9879eab2cea8",
diff --git a/Simul-Robot/c001_episode_007650/source/cameras.json b/Simul-Robot/c001_episode_007650/source/cameras.json
index aa61ec7ec4708dce26ff7d231515244d83a919ce..e3d5dff41e1b6deac97d6343d0833b7b9a077610 100644
--- a/Simul-Robot/c001_episode_007650/source/cameras.json
+++ b/Simul-Robot/c001_episode_007650/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_007650/source/meta.json b/Simul-Robot/c001_episode_007650/source/meta.json
index c2a2a67f75e3eb015ae2955a512743ccb8ce3940..0995ad1e7eb2449db552abf9d769cabf3c739499 100644
--- a/Simul-Robot/c001_episode_007650/source/meta.json
+++ b/Simul-Robot/c001_episode_007650/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_007650/source/visibility.json b/Simul-Robot/c001_episode_007650/source/visibility.json
index 999f44635d7d70fd18e14cf54541f62029af6133..9ca3d6508a381ed5c84d92830c052df4e47d48e4 100644
--- a/Simul-Robot/c001_episode_007650/source/visibility.json
+++ b/Simul-Robot/c001_episode_007650/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": "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,
@@ -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_008040/sample.json b/Simul-Robot/c001_episode_008040/sample.json
index 7982e4323ea223882623969cdf3f0d4d27de9b02..e3513370d145ebe189314dcfe5c6c580bcbfd05e 100644
--- a/Simul-Robot/c001_episode_008040/sample.json
+++ b/Simul-Robot/c001_episode_008040/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "c21e31fe95dcf6392564db3f50bedaa1a6253e7544ff283d7db93037617cd745",
   "labels/objects.json": "e2c827e32023b87c12fd2f881a043481fc7dd8dd6cad4e053add5e5b9d7f7c65",
   "labels/visibility_states.json": "281364d3c5dd44010827b440dc3a166848202aa531c8179b01dd18962d70618d",
-  "source/cameras.json": "abb6c943065dd310cb621f086c643c62de140426d3374b53ecee55de9eb371da",
+  "source/cameras.json": "6f1ca8e33fd568200a3dcf250016b6d8a071074fedf2afac1b29ce7ad69ea363",
   "source/caption.txt": "dcd5628cd7e16d054bacdab5122838f7b49d00fdfc5f0b1a64e58518662b1ea7",
-  "source/meta.json": "4cf9f7fc00140c57966db3f8d6104384f57f721985d67d0e164c3233ab8dccd6",
-  "source/points_target.npz": "ec5bc811a28060766412fcc23b82bfa359be198404f4e344cb7fa39a9d530ec0",
-  "source/visibility.json": "493672a4e3db35fac080ac25f49d62ae9be57c8d974e5861b308ac702fb8e8e4",
+  "source/meta.json": "865036c127b2f1e14677a293bf643fbb3895dc91f620d73500da7d4d9abc7fca",
+  "source/points_target.npz": "0f8f7af2550f0a757efc86ff697f915787a2d7137247e07b55e58f3da6dd05fd",
+  "source/visibility.json": "c4d81c1c4767d7eca3874605d9120bea2e867692ca13e9a727f64b8c91d50c63",
   "tracks3d.npz": "2d07b6b985f12789cda412b3e714f0e1eb25621e7d4c09a97dbad5b1d135db06",
   "videos/view0.mp4": "82d1969d1fff6836b9124ae9a539415d16669fa5d122b34a98e5ea294544fbc9",
   "videos/view1.mp4": "15ed0e6690ca0127d16b85b7933ff2e654cc04c87db2b6d24289c1262b929ec7",
diff --git a/Simul-Robot/c001_episode_008040/source/cameras.json b/Simul-Robot/c001_episode_008040/source/cameras.json
index 5775d5cbcd65eee38f6f8aa117274b308732ac20..74d1a1898a8d1fd8d4d31c87db634c10ba8bc717 100644
--- a/Simul-Robot/c001_episode_008040/source/cameras.json
+++ b/Simul-Robot/c001_episode_008040/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_008040/source/meta.json b/Simul-Robot/c001_episode_008040/source/meta.json
index 86e1d0f71ba4182b0618877c3edcb1fe23c76143..11585199361f06ab6e218253f699404c00bde96d 100644
--- a/Simul-Robot/c001_episode_008040/source/meta.json
+++ b/Simul-Robot/c001_episode_008040/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_008040/source/visibility.json b/Simul-Robot/c001_episode_008040/source/visibility.json
index 30bc986d27f39c776b179f64eb36fc31203ff9dc..170c36b1774014efd8985942a504cfbec16b6f0f 100644
--- a/Simul-Robot/c001_episode_008040/source/visibility.json
+++ b/Simul-Robot/c001_episode_008040/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_frog_can": [
@@ -4171,9 +4171,9 @@
       "target": "small_yellow_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,
@@ -7469,5 +7469,5 @@
       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_001020/sample.json b/Simul-Robot/k001_episode_001020/sample.json
index f60081dedcf8bd400ccd2fa848d4ff692923b502..f1dd0f13eba55301ece02bebf2e1d89589bdb379 100644
--- a/Simul-Robot/k001_episode_001020/sample.json
+++ b/Simul-Robot/k001_episode_001020/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "2db563cfafd6eec75a0df387d19ac967b913961996a9fadfdd3ad69c942de1ab",
   "labels/objects.json": "d98c8c5b1e88672b1b28a16271c456d7bfc2540f0b61a070fb5e102135addf6a",
   "labels/visibility_states.json": "f137eb5452295cef2c867c2838b432930e618875ba50c39f987759521a0db56e",
-  "source/cameras.json": "701a3dec9b3a8f7a2cc0d100635e4bbba48d0f01251bc4053d9d388bd8acea51",
+  "source/cameras.json": "ac7979ebeb50d21c27157d7aa92703d02307846f46a1e55f2c75849dd092e50e",
   "source/caption.txt": "b6bee126083295bd06a825c94685f29c874fcd5c141c13b550fb7f843d2b1795",
-  "source/meta.json": "5a6739791806e07accf7b38c6da78a9592dbcbef7561299fdc9e6f6ea48e8e6e",
-  "source/points_target.npz": "960e002a104dcd9c147f595c440452fb19c4dc8e1876dd28939078662531930d",
-  "source/visibility.json": "55972a85b87224fabc817f2a87805b93f01ba5d6ec73ff60fda54eef94860c13",
+  "source/meta.json": "8b3fff35a66976c9e5e13a954c9e622c40a6ce5078b642cfe8ff05686e41f522",
+  "source/points_target.npz": "413604c568b683bfcf811a7ee6e0a4b96eef0881e069dcd4973c5b3b8debe3c8",
+  "source/visibility.json": "7c341dfdb7326cb7e09bf072436d54fcb7e81cc2ed790000b63124902aed888d",
   "tracks3d.npz": "2af18546f9d444cc269b72fcce9b4f8b06729958cf8c725de682c45b5612b680",
   "videos/view0.mp4": "6926d7b92f168e245a2abf4253882b276510137b1c3eb92d3cead5c375ccf5d7",
   "videos/view1.mp4": "fa26ec6849f6af3b27f162bd015b0a369581c5cecfc46ace7e13becee37b9ba6",
diff --git a/Simul-Robot/k001_episode_001020/source/cameras.json b/Simul-Robot/k001_episode_001020/source/cameras.json
index a875c83838cd118c9af834485ad7df335e9012c7..6f9cc1cd35e7096e60425a0c7678fcf702086ea9 100644
--- a/Simul-Robot/k001_episode_001020/source/cameras.json
+++ b/Simul-Robot/k001_episode_001020/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_001020/source/meta.json b/Simul-Robot/k001_episode_001020/source/meta.json
index 54d8fc5ec69ba77ca78d6c4751ce873342ad8696..8e347f3824d5c571c2ed44159d1335ad3c8db2e6 100644
--- a/Simul-Robot/k001_episode_001020/source/meta.json
+++ b/Simul-Robot/k001_episode_001020/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/k001_episode_001020/source/visibility.json b/Simul-Robot/k001_episode_001020/source/visibility.json
index c411b6ab0ca0cc5acc708e23c95e4c70611ec7c2..6d466a8355c742a279d833b21a6aac5ac91946a3 100644
--- a/Simul-Robot/k001_episode_001020/source/visibility.json
+++ b/Simul-Robot/k001_episode_001020/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_002160/sample.json b/Simul-Robot/k002_episode_002160/sample.json
index 9f578468bcd6dbf3eddc8e53d9cbc1c539dd07ac..696e99a2c971d625d03e30a4ee047a42a02f85e5 100644
--- a/Simul-Robot/k002_episode_002160/sample.json
+++ b/Simul-Robot/k002_episode_002160/sample.json
@@ -217,11 +217,11 @@
   "depth/depth.zip": "9b0056354114f6658171dd02e34e32e54cdfd62f89e4dee453f3d05ef01dde71",
   "labels/objects.json": "7afdb81abef663e323ae907fbf619d282259e6b3d3b0fb0d2bc4b171662acefa",
   "labels/visibility_states.json": "3902884b67c8501aeca60a4b17414149388081f74ae9bba602de1fb03af96e76",
-  "source/cameras.json": "bc08acf6debf63414de1cf7a2ff0a5d15db07317b2502d512bb56ad8d28adeab",
+  "source/cameras.json": "515fc6bd6dcceb52fdae25c74c6dfbe76d820c6b900bf6ef2d165de4f4bcacbf",
   "source/caption.txt": "77103e7906fa5964dd0e5826b2ee50ad6be4845d15703a1f593d383dccc7f53d",
-  "source/meta.json": "873b26de13437b17256c0f7846beb6773d60664529661052fe2376a306c55691",
-  "source/points_target.npz": "b5bb7d718a634614a9f6cf83e7bd0b6389b5627cfd60c27b684975eb36c217db",
-  "source/visibility.json": "cfd654a974bad13d64c06f5453f72cb6cece0d921041961108de38c86bb6aa66",
+  "source/meta.json": "a5dd83a0a06c116d9e062d48c067f230b9046144bd0e6b353042ad5f1db264ac",
+  "source/points_target.npz": "49334084556923250f39ee17f9fe0a5a3c3e967f95c5376ec559a08b80aec777",
+  "source/visibility.json": "03a28353c8422f7b9c99e2457642c4d574aa97f68f871c674cd71ba90e299a98",
   "tracks3d.npz": "89f0eea99083c7e49c58445e3e1f14986ab6695841f95df7893bb4894f74a17c",
   "videos/view0.mp4": "12c81c520a283db9eadd9166896340824140e444b7cbeee970d3f6c54ee05a21",
   "videos/view1.mp4": "8ae1e6bd4aa0f4a66fbeceb791ee1e4d7d23409a1e691a85ba1febce8e1bdda8",
diff --git a/Simul-Robot/k002_episode_002160/source/cameras.json b/Simul-Robot/k002_episode_002160/source/cameras.json
index 8a9f524b764717c71eb9c17ef2db4148cf16494f..bbe281333453dbe4c0e27596401ed813d9f10a11 100644
--- a/Simul-Robot/k002_episode_002160/source/cameras.json
+++ b/Simul-Robot/k002_episode_002160/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_002160/source/meta.json b/Simul-Robot/k002_episode_002160/source/meta.json
index 47a6eb9949d63a9d7209e0bd6831361165b1ba7b..3cc70750ea5f74c5b300d160e7d80344f07d3667 100644
--- a/Simul-Robot/k002_episode_002160/source/meta.json
+++ b/Simul-Robot/k002_episode_002160/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/k002_episode_002160/source/visibility.json b/Simul-Robot/k002_episode_002160/source/visibility.json
index 4bd7452222ca262d00b3ad1f1f4c93493cb14831..0028222f57d349faa97695e17cedae87d197ac17 100644
--- a/Simul-Robot/k002_episode_002160/source/visibility.json
+++ b/Simul-Robot/k002_episode_002160/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": [
@@ -5443,9 +5443,9 @@
       "target": "yellow_dog_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,
@@ -9053,5 +9053,5 @@
       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