domino_chain: retime the timed DELETE to frame 24 (mid-fall)

edit_remove_domino_1_after_cascade (DELETE domino_1 AT FRAME 48) is replaced by edit_remove_domino_1_mid_fall (AT FRAME 24). At frame 24 the first tile is still falling at 38 deg, three frames after it first touches tile 2, so the cascade still completes but lags the source by 6-8 frames. 132 edits / 152 clips.

README.md

@@ -19,7 +19,7 @@ size_categories:
 # PCVE-RigidBench
 
 **P**hysics-**C**onsistent **V**ideo **E**diting benchmark, **rigid-body** subset:
-20 scenes, 20 source videos, 131 edit tasks. Each task asks a model to re-render
+20 scenes, 20 source videos, 132 edit tasks. Each task asks a model to re-render
 a source video under a changed physical parameter (mass, friction, restitution,
 initial velocity, or the presence of an object) and provides the
 physically-simulated ground truth to score against.
@@ -29,7 +29,7 @@ every case ships exact per-frame object trajectories alongside the video.
 
 ## Scope: rigid bodies only
 
-All 151 clips are **rigid-body** simulations. Every dynamic object is a rigid
+All 152 clips are **rigid-body** simulations. Every dynamic object is a rigid
 body in PyBullet, and every interaction is a rigid contact parameterised by
 mass, friction, rolling friction and restitution. Static scene geometry (floors,
 tables, ramps, the rug in `ball_carpet_climb`) participates purely as collision
@@ -95,9 +95,9 @@ per-frame object trajectories, or against `edited_video` for pixel /
 perceptual metrics.
 
 ## Counts
-- Edits by kind: {'SET': 92, 'DELETE': 32, 'ADD': 7}
-- Edits by timing: {'partway': 14, 'whole_clip': 117}
-- Edits by property: {'friction': 21, 'initial_velocity': 26, 'mass': 32, 'presence': 39, 'restitution': 13}
+- Edits by kind: {'SET': 92, 'DELETE': 33, 'ADD': 7}
+- Edits by timing: {'partway': 15, 'whole_clip': 117}
+- Edits by property: {'friction': 21, 'initial_velocity': 26, 'mass': 32, 'presence': 40, 'restitution': 13}
 
 ## Notes
 - `ground_truth_transforms.json` is the physics record, not a render log: the

benchmark_manifest.json

@@ -42,21 +42,21 @@
   "counts": {
     "scenes": 20,
     "sources": 20,
-    "edits": 131,
+    "edits": 132,
     "by_property": {
       "friction": 21,
       "initial_velocity": 26,
       "mass": 32,
-      "presence": 39,
+      "presence": 40,
       "restitution": 13
     },
     "by_kind": {
       "SET": 92,
-      "DELETE": 32,
+      "DELETE": 33,
       "ADD": 7
     },
     "by_timing": {
-      "partway": 14,
+      "partway": 15,
       "whole_clip": 117
     }
   },
@@ -3743,6 +3743,49 @@
       "ground_truth": "scenes/domino_chain/cases/edit_remove_domino_1/ground_truth_transforms.json",
       "thumbnail": "thumbnails/domino_chain/edit_remove_domino_1.jpg"
     },
+    {
+      "global_id": "domino_chain/edit_remove_domino_1_mid_fall",
+      "scene": "domino_chain",
+      "case_id": "edit_remove_domino_1_mid_fall",
+      "source_case_id": "domino_chain_baseline",
+      "source_video": "scenes/domino_chain/cases/domino_chain_baseline/video.mp4",
+      "edited_video": "scenes/domino_chain/cases/edit_remove_domino_1_mid_fall/video.mp4",
+      "duration_sec": 4.0,
+      "prompts": {
+        "vague": {
+          "zh": "在视频前段,把第 1 张骨牌从场景中删除。",
+          "en": "Early in the clip, remove the first domino from the scene."
+        },
+        "quantitative": {
+          "zh": "从第 24 帧开始,把第 1 张骨牌从场景中删除。",
+          "en": "From frame 24 onwards, remove the first domino from the scene."
+        }
+      },
+      "edit_dsl": "DELETE domino_1 AT FRAME 24",
+      "edit_kind": "DELETE",
+      "object_id": "domino_1",
+      "property": null,
+      "outcome_type": "object_removed",
+      "applies_from_frame": 24,
+      "timing": "partway",
+      "physics_diff": {
+        "domino_1": {
+          "from": "present",
+          "to": "removed"
+        },
+        "timing": {
+          "applies_from_frame": 24,
+          "total_frames": 96,
+          "unedited_frames": [
+            1,
+            23
+          ]
+        }
+      },
+      "edit_summary": "First tile removed at frame 24 of 96, while it is still falling. Frames 1-23 are the source video frame for frame: tile 1 tips from its 12 degree start and reaches 38 degrees, having met tile 2 at frame 21 and started it moving (tile 2 is 5 degrees over at frame 24). Tile 1 then vanishes mid-fall, leaning but nowhere near down, and the rest of the row carries on under the momentum it already had: tiles 2, 3 and 4 all still topple, but each crosses 45 degrees six to eight frames later than in the source (frames 43, 51, 55 against 37, 43, 47) and the row settles around frame 60 instead of frame 52. The end state matches the baseline -- 76, 76 and flat at 90 degrees, within 0.2 degrees and 4 cm -- so what separates this from the source is the missing tile plus a visibly slower cascade. Toppled=3/4. The whole-clip version of the same delete (edit_remove_domino_1) is a different video again: with no pre-tilted tile to start it, nothing in the row ever moves.",
+      "ground_truth": "scenes/domino_chain/cases/edit_remove_domino_1_mid_fall/ground_truth_transforms.json",
+      "thumbnail": "thumbnails/domino_chain/edit_remove_domino_1_mid_fall.jpg"
+    },
     {
       "global_id": "domino_chain/edit_remove_domino_2",
       "scene": "domino_chain",

scenes/domino_chain/cases/edit_remove_domino_1_mid_fall/prompts.json

@@ -0,0 +1,32 @@
+{
+  "schema_version": 2,
+  "case_id": "edit_remove_domino_1_mid_fall",
+  "source_case_id": "domino_chain_baseline",
+  "edit_dsl": "DELETE domino_1 AT FRAME 24",
+  "edit_summary": "First tile removed at frame 24 of 96, while it is still falling. Frames 1-23 are the source video frame for frame: tile 1 tips from its 12 degree start and reaches 38 degrees, having met tile 2 at frame 21 and started it moving (tile 2 is 5 degrees over at frame 24). Tile 1 then vanishes mid-fall, leaning but nowhere near down, and the rest of the row carries on under the momentum it already had: tiles 2, 3 and 4 all still topple, but each crosses 45 degrees six to eight frames later than in the source (frames 43, 51, 55 against 37, 43, 47) and the row settles around frame 60 instead of frame 52. The end state matches the baseline -- 76, 76 and flat at 90 degrees, within 0.2 degrees and 4 cm -- so what separates this from the source is the missing tile plus a visibly slower cascade. Toppled=3/4. The whole-clip version of the same delete (edit_remove_domino_1) is a different video again: with no pre-tilted tile to start it, nothing in the row ever moves.",
+  "applies_from_frame": 24,
+  "physics_diff": {
+    "domino_1": {
+      "from": "present",
+      "to": "removed"
+    },
+    "timing": {
+      "applies_from_frame": 24,
+      "total_frames": 96,
+      "unedited_frames": [
+        1,
+        23
+      ]
+    }
+  },
+  "prompts": {
+    "vague": {
+      "zh": "在视频前段,把第 1 张骨牌从场景中删除。",
+      "en": "Early in the clip, remove the first domino from the scene."
+    },
+    "quantitative": {
+      "zh": "从第 24 帧开始,把第 1 张骨牌从场景中删除。",
+      "en": "From frame 24 onwards, remove the first domino from the scene."
+    }
+  }
+}

scenes/domino_chain/cases/edit_remove_domino_1_mid_fall/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:598d601d399389ccb5d1d300aa05283272567190bfbcad798157883dca3fc032
+size 243357

splits/all.txt

@@ -66,6 +66,7 @@ dining_chain/edit_remove_cup_mid_slide
 domino_chain/edit_heavy_domino_2
 domino_chain/edit_heavy_domino_3
 domino_chain/edit_remove_domino_1
+domino_chain/edit_remove_domino_1_mid_fall
 domino_chain/edit_remove_domino_2
 domino_chain/edit_remove_domino_3
 domino_chain/edit_remove_domino_4

splits/by_property/presence.txt

@@ -13,6 +13,7 @@ dining_chain/edit_remove_cup
 dining_chain/edit_remove_milk
 dining_chain/edit_remove_cup_mid_slide
 domino_chain/edit_remove_domino_1
+domino_chain/edit_remove_domino_1_mid_fall
 domino_chain/edit_remove_domino_2
 domino_chain/edit_remove_domino_3
 domino_chain/edit_remove_domino_4

splits/by_timing/partway.txt

@@ -5,6 +5,7 @@ bouncing_ball/edit_dead_ball_after_first_bounce
 car_gap_jump/edit_grippy_wheels_mid_jump
 car_ramp_climb/edit_grippy_wheels_mid_climb
 dining_chain/edit_remove_cup_mid_slide
+domino_chain/edit_remove_domino_1_mid_fall
 picnic_apple_ball/edit_remove_apple_after_impact
 pool_collision/edit_remove_target_ball_after_impact
 ramp_collision/edit_grippy_red_ball_mid_roll