Datasets:

AI-YesLab
/

3DVerifierBench

@@ -28,58 +28,63 @@ Shuanglin Li, Jiaying Lin, Shuquan Ye.
 [Paper: arXiv:2610.22360](https://arxiv.org/abs/2610.22360)
-The test set contains **2,581 examples across ten tasks**: 970 True and 1,611
-False. The package includes original model-visible images/videos and the available
-source/candidate 3D assets needed to inspect and reconstruct the benchmark cases.
 ## Directory layout
-Each task is organized into numbered case groups. A case contains the related
-examples from one original family, their source/candidate assets and a separate
-folder of visual inputs for each example:
 ```text
 3DVerifierBench/
 ├── README.md
 ├── metadata/
 │   ├── benchmark.jsonl
-│   ├── dataset_stats.json
 │   ├── inputs.jsonl
 │   ├── motion_frames16_inputs.jsonl
-│   └── sample_asset_links.jsonl
 ├── evidence/
-│   ├── object/
-│   │   └── case_0001/
-│   │       ├── source.glb
-│   │       ├── candidate_01.glb
-│   │       ├── sample_01/
-│   │       │   ├── view_01.png
-│   │       │   └── ...
-│   │       └── sample_02/
-│   ├── part/
-│   ├── material/
-│   ├── part_grounding/
-│   │   └── case_0001/
-│   │       ├── source.glb
-│   │       ├── candidate_01.glb
-│   │       ├── candidate_02.glb
-│   │       └── sample_01/
-│   │           ├── views.png
-│   │           ├── annotated_global.png
-│   │           └── annotated_detail.png
-│   ├── part_relation/
-│   ├── affordance_region/
-│   ├── grasp/
-│   │   └── case_0001/
-│   │       ├── source.glb
-│   │       └── sample_01/
-│   │           ├── scene.glb
-│   │           ├── views.png
-│   │           ├── annotated_global.png
-│   │           └── annotated_detail.png
-│   ├── identity/
-│   ├── fine_geometry/
-│   └── motion/
 ├── manifests/
 │   ├── benchmark_manifest.json
 │   ├── pair_links.jsonl
@@ -87,111 +92,76 @@ folder of visual inputs for each example:
 └── licenses/
 ```
-There are 1,167 case groups. Original sample IDs remain in
-`metadata/benchmark.jsonl`; `display_id`, `case_directory` and `sample_directory`
-provide the readable folder mapping. Every evidence path is relative to the
-dataset root. Directory and file names use neutral identifiers.
-Each sample's `assets3d` list declares the exact source/candidate asset paths and
-their roles. Related samples share case-level meshes. Some cases contain multiple
-candidate meshes; use the sample's declared path rather than guessing from its
-number. Images are stored separately for each sample for direct browsing.
-## Visual inputs and 3D source formats
-- Core tasks use eight ordered files, `view_01.png` through `view_08.png`.
-- Region, relation and grasp tasks use `views.png`, `annotated_global.png` and
-  `annotated_detail.png`.
-- Every Grasp sample also contains `scene.glb`: the original object with the
-  candidate palm and both fingers placed at that sample's recorded pose.
 - Identity and FineGeo use `pair.png`.
-- Motion uses `video.mp4` or the alternative `frames16.png`. Its original
-  OBJ frames are expanded as `mesh_sequence/frame_00001.obj`, with `captions.txt`.
-- Identity assets include `identity_a.obj`/`.ply`, `identity_b.obj`/`.ply`, source
-  view folders and synthesis parameters.
-- FineGeo includes the supplied candidate GLBs where available, reference
-  previews and official `reference_cad.ply` files. The reference CADs retain their
-  upstream units; `metadata/fine_geometry_models_info.json` provides their metadata.
-The layout contains 11,769 image/video input files representing 7,681 unique
-original contents, plus 12,420 source asset files, including 8,745 directly stored
-GLB/OBJ/PLY meshes, and 200 derived Grasp scene GLBs. The 134 Motion sequences
-contain 6,088 original OBJ frames. Original file bytes, evidence order, claims,
-captions, labels and reference explanations are preserved.
-Forty DB-3DME examples have original source turntable GIFs rather than meshes.
-They retain `source_turntable.gif` and `geometry_status: source_turntable_only`.
-Thirty FineGeo examples have the official reference CAD without an original
-candidate mesh; their status is `official_reference_cad_only`. No missing mesh is
-invented. Exact provided visual inputs remain authoritative for reproducing
-the paper's image/video evaluation; a new render may use different cameras/pixels.
-Grasp `source.glb` contains the shared object alone. Each sample's `scene.glb`
-is self-contained and combines that unchanged object geometry with the original
-three-box parallel-jaw gripper geometry, dimensions and pose. It is declared in
-`assets3d` with role `grasp_scene`. Nodes are named `object`, `gripper_root`,
-`palm`, `left_finger` and `right_finger`. Valid and invalid candidate poses are
-both preserved, including two recorded zero-separation grippers.
-The original transforms and gripper parameters remain in
-`metadata/grasp_poses.jsonl`, matched by sample ID. Poses use source coordinates
-in meters, before object canonicalization. The complete scene applies the same
-canonicalization to the object and gripper, then rotates both from the source
-Z-up frame to glTF Y-up. Rebuild all 200 scenes from the code release with:
 ```bash
 python scripts/export_grasp_scenes.py --benchmark-root /path/to/3DVerifierBench
 ```
-This export requires only Python's standard library. Scene assets are for 3D
-inspection and reconstruction; the original supplied images remain the inputs
-for the paper's image-based Grasp evaluation.
 ## Load
-Load all test annotations directly from the Hub:
 ```python
 from datasets import load_dataset
 test = load_dataset("AI-YesLab/3DVerifierBench", split="test")
 ```
-Download the complete repository, including the images, videos and 3D assets,
-to resolve each sample's relative evidence and asset paths locally:
 ```python
 from pathlib import Path
 from huggingface_hub import snapshot_download
 root = Path(snapshot_download(repo_id="AI-YesLab/3DVerifierBench", repo_type="dataset"))
 ```
-For an already downloaded or extracted dataset, load the local annotation file:
 ```python
-from pathlib import Path
-from datasets import load_dataset
-root = Path("/path/to/3DVerifierBench")
-test = load_dataset(
-    "json", data_files={"test": str(root / "metadata/benchmark.jsonl")}, split="test"
-)
-sample = test[0]
-views = [root / item["path"] for item in sample["evidence"]["items"]]
-assets = [root / item["path"] for item in sample["assets3d"]]
 ```
-Install the reader with `python -m pip install datasets`. The default benchmark
-configuration declares the `test` split. `metadata/inputs.jsonl` contains target-free
-subject inputs; Motion frame alternatives are in `metadata/motion_frames16_inputs.jsonl`.
-The supplied image/video input protocol does not pass source meshes, labels,
-reference explanations or sibling links into the subject prompt.
 ## Evaluate
-Install the separate `3DVerifier` code release and configure your model endpoint,
-model name and API key as documented there. From that code checkout:
 ```bash
 python scripts/prepare_manifest.py --benchmark-root /path/to/3DVerifierBench --task core.part --output work/part_manifest
@@ -199,22 +169,23 @@ python scripts/run_evaluation.py --manifest work/part_manifest --config examples
 python scripts/score_run.py --run runs/part
 ```
-Native Motion uses the video transport entry point. Preserve each model's original
-native-video or frame-sheet protocol for paper reproduction.
-## Manifests and licensing
-`manifests/benchmark_manifest.json` records membership, paths, file sizes and
-hashes for visual inputs and source assets. `metadata/dataset_stats.json` gives
-per-task counts. Linked evaluation retains 661 Spec pairs, 727 Candidate pairs
-and 615 triplets. The benchmark's 1,123 recorded source groups are disjoint from
-the training resource's 4,011 groups; unrecorded semantic near-duplicates are
-outside this grouping check.
-Original annotations are CC BY 4.0. Upstream images, videos, meshes, captions and
-native annotations retain their source and per-asset terms. See
-[licenses and attribution](licenses/README.md) and retain the included notices.
-Model weights and training examples are separate distributions.
 Please cite the [arXiv preprint](https://arxiv.org/abs/2610.22360):

 [Paper: arXiv:2610.22360](https://arxiv.org/abs/2610.22360)
+The test set contains **2,581 samples across ten tasks**, with 970 True and
+1,611 False answers. It includes the original image/video inputs and available
+3D assets. Related samples are grouped into 1,167 cases.
+## Test annotations
+`metadata/benchmark.jsonl` contains one sample per line, with eight fields:
+| Field | Meaning |
+| --- | --- |
+| `sample_id` | Short neutral ID, such as `object_0001_01` |
+| `task_id` | Evaluation task |
+| `question` | The factual claim; for Motion, the complete caption-based instruction |
+| `answer` | Ground-truth Boolean judgment |
+| `reference_explanation` | Reference rejection explanation used for scoring |
+| `instruction` | Static-task instruction; Motion's instruction is already in `question` |
+| `input_type` | Original image/video input protocol |
+| `evidence` | Ordered image/video entries with paths, media IDs and content hashes |
+IDs follow `<task>_<case>_<sample>`, matching the evidence directory. For example,
+`grasp_0090_01` uses `evidence/grasp/case_0090/sample_01/`. Sample numbers use a
+fixed neutral ordering and do not encode the answer. Each sample is evaluated
+independently. All paths are relative to the dataset root.
+`metadata/inputs.jsonl` is the same public input projection without answers or
+reference explanations. Core tasks use the question and eight ordered views;
+their generic task instruction is omitted from model input, following the paper.
+`metadata/motion_frames16_inputs.jsonl` supplies the 134 alternative sixteen-frame
+Motion inputs with the original sampling and layout information.
+Evaluation grouping, sample construction type, source attribution and geometry
+availability are in `metadata/sample_metadata.jsonl`. The exact source/candidate
+3D asset associations are in `metadata/sample_asset_links.jsonl`. Join these
+indexes by the same short `sample_id`; asset filenames have independent numbering.
+The code release's reader joins the indexes automatically. Original questions,
+answers, reference explanations, media bytes and evidence order are preserved.
 ## Directory layout
 ```text
 3DVerifierBench/
 ├── README.md
 ├── metadata/
 │   ├── benchmark.jsonl
 │   ├── inputs.jsonl
 │   ├── motion_frames16_inputs.jsonl
+│   ├── sample_metadata.jsonl
+│   ├── sample_asset_links.jsonl
+│   ├── grasp_poses.jsonl
+│   └── dataset_stats.json
 ├── evidence/
+│   └── <task>/case_0001/
+│       ├── source.glb
+│       ├── candidate_01.glb
+│       └── sample_01/
+│           ├── view_01.png
+│           └── ...
 ├── manifests/
 │   ├── benchmark_manifest.json
 │   ├── pair_links.jsonl
 └── licenses/
 ```
+The tasks are Object, Part, Material, Part Grounding, Part Relation, Affordance
+Region, Grasp Validity, Identity Preservation, Fine-grained Geometry and Motion.
+## Images, videos and 3D assets
+- Core tasks have eight ordered images, `view_01.png` through `view_08.png`.
+- Region, relation, affordance and Grasp use `views.png`, `annotated_global.png`
+  and `annotated_detail.png`.
 - Identity and FineGeo use `pair.png`.
+- Motion has `video.mp4`, the alternative `frames16.png`, and original OBJ frames
+  under `mesh_sequence/frame_00001.obj`, with `captions.txt`.
+- Identity assets include A/B OBJ and PLY meshes, source views and synthesis data.
+- FineGeo includes supplied candidate GLBs and official `reference_cad.ply` files.
+  `metadata/fine_geometry_models_info.json` records their upstream CAD metadata.
+There are 11,769 image/video input files, representing 7,681 unique original
+contents, plus 12,420 source asset files and 200 complete Grasp scene GLBs. The
+134 Motion sequences contain 6,088 original OBJ frames.
+Every Grasp sample contains `scene.glb` with its unchanged object, palm and both
+fingers at the recorded candidate pose. Its case-level `source.glb` remains the
+object alone. `metadata/grasp_poses.jsonl` retains the 4x4 source transform and
+gripper dimensions in meters. The complete scene applies the same canonicalization
+to object and gripper, then converts source Z-up to glTF Y-up. Rebuild the scenes
+from the code checkout with:
 ```bash
 python scripts/export_grasp_scenes.py --benchmark-root /path/to/3DVerifierBench
 ```
+Forty DB-3DME samples have original turntable GIFs instead of source meshes.
+Thirty FineGeo samples have official reference CADs without an original candidate
+mesh. Their `geometry_status` is recorded in `metadata/sample_metadata.jsonl`.
+Provided images/videos remain authoritative for reproducing the paper's visual
+input protocol; new renders may have different cameras and pixels.
 ## Load
 ```python
 from datasets import load_dataset
 test = load_dataset("AI-YesLab/3DVerifierBench", split="test")
+print(test[0]["sample_id"], test[0]["question"], test[0]["answer"])
 ```
+Download the complete repository to resolve evidence and 3D asset paths locally:
 ```python
 from pathlib import Path
 from huggingface_hub import snapshot_download
 root = Path(snapshot_download(repo_id="AI-YesLab/3DVerifierBench", repo_type="dataset"))
+sample = test[0]
+views = [root / item["path"] for item in sample["evidence"]]
 ```
+For an already extracted ZIP, use the local JSON reader:
 ```python
+test = load_dataset("json", data_files={"test": str(root / "metadata/benchmark.jsonl")}, split="test")
 ```
+Install the reader with `python -m pip install datasets`. Source meshes, labels,
+reference explanations and related-sample metadata are excluded from subject
+prompts. The separate `3DVerifier` code release handles prompt construction,
+model execution and scoring.
 ## Evaluate
+From the code checkout, configure your own model endpoint and credentials:
 ```bash
 python scripts/prepare_manifest.py --benchmark-root /path/to/3DVerifierBench --task core.part --output work/part_manifest
 python scripts/score_run.py --run runs/part
 ```
+For native Motion, prepare its video manifest with
+`scripts/prepare_motion_manifest.py` and use `scripts/evaluate_motion_video.py`.
+Preserve the model's original native-video or sixteen-frame protocol for paper
+reproduction.
+## Indexes, license and citation
+`manifests/benchmark_manifest.json` records paths, membership and content hashes.
+`metadata/dataset_stats.json` provides global and per-task counts. The linked
+indexes contain 661 specification pairs, 727 candidate pairs and 615 triplets.
+The benchmark's 1,123 recorded source groups are disjoint from the training
+resource's 4,011 groups; this check does not cover unrecorded semantic duplicates.
+Original annotations are CC BY 4.0. Upstream media, meshes, captions and native
+annotations retain their source-specific terms. See
+[licenses and attribution](licenses/README.md) and retain the supplied notices.
+Training data and model weights are separate distributions.
 Please cite the [arXiv preprint](https://arxiv.org/abs/2610.22360):

manifests/benchmark_manifest.json CHANGED Viewed

@@ -1,5 +1,5 @@
 {
-  "schema_version": "3dverifierbench-benchmark-manifest-v1",
   "dataset": "3DVerifierBench",
   "version": "1.0.0",
   "split": "test",
@@ -14,43 +14,47 @@
   "evidence_root": ".",
   "media_file_count": 11769,
   "unique_media_contents": 7681,
-  "sample_ids_sha256": "43e9e17d1100fec8e990e14ceaecc8e6db71e5ff32a57add1cacdc371e180d27",
   "annotation_files": {
     "manifests/pair_links.jsonl": {
-      "sha256": "66c6fb914bd659ff592d0e140cc27694e8884dee7c0caa779c426496714ccf52",
-      "bytes": 625969
     },
     "manifests/triplet_links.jsonl": {
-      "sha256": "6bef484e85037f7c3b57165a72e474a989f368543007a6749deede99adb454d4",
-      "bytes": 372037
     },
     "metadata/benchmark.jsonl": {
-      "sha256": "11e3c33021fb7eacd28979e7dcb7ec4a93ca8b16895b04a150f95f5a3fd93454",
-      "bytes": 7265842
     },
     "metadata/dataset_stats.json": {
       "sha256": "bf004cc513904681f23cb6ab34d29a82c5852a4c4c8ee247e2ed50d242ce9ce1",
       "bytes": 4155
     },
     "metadata/grasp_poses.jsonl": {
-      "sha256": "38c239ea0596b5f92d08fb2a88c91ec47f37b5e3bcbe3f09899d97c8dbd8bdb6",
-      "bytes": 172030
     },
     "metadata/inputs.jsonl": {
-      "sha256": "b635f822ba9d611bdffb331fa7674fd8371c36306ba3b49b407088d8057a0ab0",
-      "bytes": 3616709
     },
     "metadata/motion_frames16_inputs.jsonl": {
-      "sha256": "dad398fd0d46caf1535b751f3111f6aad8a482ffc2aaff796ae36298a96e6cb7",
-      "bytes": 160820
     },
     "metadata/non_mesh_source_cases.json": {
-      "sha256": "4b7e028b85fe1e9e6477e13217d38f1b014b7ad933d36200d38868a83ffd0332",
-      "bytes": 7063
     },
     "metadata/sample_asset_links.jsonl": {
-      "sha256": "4afd3fb605305559016d78d1c1c60d1f78e26d5bc9dae435dbcf7df3c5c09f57",
-      "bytes": 3268983
     }
   },
   "media_files": {
@@ -202627,5 +202631,16 @@
   "case_groups": 1167,
   "animation_sequences": 134,
   "animation_obj_frames": 6088,
-  "derived_grasp_scene_count": 200
 }

 {
+  "schema_version": "3dverifierbench-benchmark-manifest-v2",
   "dataset": "3DVerifierBench",
   "version": "1.0.0",
   "split": "test",
   "evidence_root": ".",
   "media_file_count": 11769,
   "unique_media_contents": 7681,
+  "sample_ids_sha256": "51241351c2cc292b75bc36bd7b6cdffc31a6bb4437a0330d5f7ce38b2cd8852f",
   "annotation_files": {
     "manifests/pair_links.jsonl": {
+      "sha256": "eef494c1d57de341b5b867b9b6cf14c39ba666bcd5324401d3c46331afe968d6",
+      "bytes": 285072
     },
     "manifests/triplet_links.jsonl": {
+      "sha256": "b69f987f95d0f41babd0b8d05c6d26d36540d25ae34602f494311cb59fd00daa",
+      "bytes": 143924
     },
     "metadata/benchmark.jsonl": {
+      "sha256": "d813406689e3c14976e2b7c9c048278c5e1043602ec5d2f1f719bb9705b2e209",
+      "bytes": 4038773
     },
     "metadata/dataset_stats.json": {
       "sha256": "bf004cc513904681f23cb6ab34d29a82c5852a4c4c8ee247e2ed50d242ce9ce1",
       "bytes": 4155
     },
     "metadata/grasp_poses.jsonl": {
+      "sha256": "617afac7f152a51c4c872ea6cad50492a554d8e13a37396a4830a847b5aa3915",
+      "bytes": 166030
     },
     "metadata/inputs.jsonl": {
+      "sha256": "b66e08b43c3de0fbce73839d417b412908c4708cfdd8b941769ea9912f77f44a",
+      "bytes": 3482776
     },
     "metadata/motion_frames16_inputs.jsonl": {
+      "sha256": "9e64f69de0b89147ce80d89e23f31c93bde156dec35948a5fd60b191b9b20ed3",
+      "bytes": 146080
     },
     "metadata/non_mesh_source_cases.json": {
+      "sha256": "2f09a2e84f9d40eaafcc817ccce848cff7109d04aabbdbb37d7da23a85cf4285",
+      "bytes": 4723
     },
     "metadata/sample_asset_links.jsonl": {
+      "sha256": "d426d98ca24cb02c675b2ed7ba917ec7b5a3bf37397a940bc40aebea8e72c48e",
+      "bytes": 2903223
+    },
+    "metadata/sample_metadata.jsonl": {
+      "sha256": "4d1bbe6144d29df12cd395d7b1edc23ed893c31713460208726cc5796436a0ee",
+      "bytes": 744427
     }
   },
   "media_files": {
   "case_groups": 1167,
   "animation_sequences": 134,
   "animation_obj_frames": 6088,
+  "derived_grasp_scene_count": 200,
+  "sample_metadata": "metadata/sample_metadata.jsonl",
+  "public_sample_schema": [
+    "sample_id",
+    "task_id",
+    "question",
+    "answer",
+    "reference_explanation",
+    "instruction",
+    "input_type",
+    "evidence"
+  ]
 }

manifests/pair_links.jsonl CHANGED Viewed