Publish SimCode4D data card, portable runtime, and scene previews

README.md

@@ -2,11 +2,15 @@
 pretty_name: SimCode4D DEV15
 language:
   - en
+size_categories:
+  - n<1K
 tags:
-  - embodied-ai
   - robotics
+  - embodied-ai
   - mujoco
+  - blender
   - benchmark
+  - 3d
 configs:
   - config_name: dev15
     data_files:
@@ -14,85 +18,92 @@ configs:
         path: data/dev15.jsonl
 ---
 
-# SimCode4D: Code2Worlds-2 development benchmark
+# SimCode4D DEV15
+
+SimCode4D is the 15-scene **development benchmark** for [Code2Worlds-2](https://github.com/AIGeeksGroup/Code2Worlds-2). It spans indoor, industrial, natural outdoor, service, and urban worlds (three scenes per domain). The table in Data Studio is a searchable scene index, not a leaderboard or a set of model outputs.
+
+This repository holds **benchmark data**: prompts, 15 portable MuJoCo scene models with required mesh assets, 12 archived Blender authoring projects, and 45 real scene previews. The benchmark evaluator, submission API, validators, and metric implementations live in the [Code2Worlds-2 GitHub repository](https://github.com/AIGeeksGroup/Code2Worlds-2/tree/main/benchmark/simcode4d_unified).
+
+## What is included
+
+| Domain | Scenes | Preview gallery |
+| --- | ---: | --- |
+| Indoor | 3 | [Three views per scene](https://huggingface.co/datasets/AIGeeksGroup/SimCode4D/resolve/main/previews/IND_contact_sheet.jpg) |
+| Industrial | 3 | [Three views per scene](https://huggingface.co/datasets/AIGeeksGroup/SimCode4D/resolve/main/previews/INS_contact_sheet.jpg) |
+| Natural outdoor | 3 | [Three views per scene](https://huggingface.co/datasets/AIGeeksGroup/SimCode4D/resolve/main/previews/NAT_contact_sheet.jpg) |
+| Service | 3 | [Three views per scene](https://huggingface.co/datasets/AIGeeksGroup/SimCode4D/resolve/main/previews/SER_contact_sheet.jpg) |
+| Urban | 3 | [Three views per scene](https://huggingface.co/datasets/AIGeeksGroup/SimCode4D/resolve/main/previews/URB_contact_sheet.jpg) |
+
+## Preview gallery
+
+### Indoor
+
+![Indoor: three previews per scene](https://huggingface.co/datasets/AIGeeksGroup/SimCode4D/resolve/main/previews/IND_contact_sheet.jpg)
+
+### Industrial
+
+![Industrial: three previews per scene](https://huggingface.co/datasets/AIGeeksGroup/SimCode4D/resolve/main/previews/INS_contact_sheet.jpg)
 
-SimCode4D DEV15 is the public, portable 15-case development benchmark for the
-Code2Worlds-2 paper. It covers indoor, industrial, natural outdoor, service,
-and urban environments, with three cases in each domain. The benchmark tests
-whether a generated world builds and whether its physical behavior satisfies
-the registered task, integrity, mechanics, kinematics, contact, friction, and
-safety checks.
+### Natural outdoor
 
-The `validation` split in the viewer is a **case index**. Each row points to
-the complete frozen specification, gold predicates, physical profile, task
-confirmation, and runtime artifacts in this repository. It is not a table of
-model outputs or experiment results.
+![Natural outdoor: three previews per scene](https://huggingface.co/datasets/AIGeeksGroup/SimCode4D/resolve/main/previews/NAT_contact_sheet.jpg)
 
-## Contents
+### Service
+
+![Service: three previews per scene](https://huggingface.co/datasets/AIGeeksGroup/SimCode4D/resolve/main/previews/SER_contact_sheet.jpg)
+
+### Urban
+
+![Urban: three previews per scene](https://huggingface.co/datasets/AIGeeksGroup/SimCode4D/resolve/main/previews/URB_contact_sheet.jpg)
+
+Each scene has three images under `previews/<case_id>/`. The natural outdoor and service previews are three frames from existing Blender animations (start, middle, end). The indoor, industrial, and urban previews are three offscreen MuJoCo views of the registered runtime world. To expose the scene interior for inspection, the visualization hides roof geometry; the saved simulation model is unchanged. These images inspect appearance and layout; they are not evidence of task success.
 
 | Path | Contents |
 | --- | --- |
-| `data/dev15.jsonl` | Searchable index of all 15 development cases |
-| `benchmark/simcode4d_unified/config/` | Suite registry, seed protocol, and frozen hashes |
-| `benchmark/simcode4d_unified/cases/` | Case specifications, gold predicates, physics profiles, and task confirmations |
-| `benchmark/simcode4d_unified/artifacts/` | Frozen M1 and executable MuJoCo runtime models and overlays |
-| `benchmark/simcode4d_unified/assets/` and `vendor/` | Runtime meshes and third-party model assets |
-| `benchmark/simcode4d_unified/probes/`, `validators/`, and `tools/` | Reference probes, independent validators, and scoring tools |
-| `requirements-benchmark.txt` | Python dependencies for the public benchmark runtime |
+| `data/dev15.jsonl` | 15 prompts, domain and task labels, asset paths, and preview paths |
+| `runtime/simcode4d_dev15_runtime.tar.gz` | Portable scene XML and required mesh assets for all 15 cases |
+| `blender_sources/<case_id>/scene.blend` | 12 available authoring projects |
+| `previews/<case_id>/` | Three rendered images per scene |
+| `THIRD_PARTY_NOTICES.md` | Asset provenance and third-party notices |
 
-The runtime package is copied from the public
-[Code2Worlds-2 repository](https://github.com/AIGeeksGroup/Code2Worlds-2/tree/ca35f25cda19aadf1d211ae3e55e390905507f1d/benchmark/simcode4d_unified)
-at commit `ca35f25cda19aadf1d211ae3e55e390905507f1d`. Paths are preserved so
-that the original registry, hashes, and evaluator work without rewriting the
-cases. The case index is an additional convenience file.
+The original `.blend` sources for **IND-01**, **IND-02**, and **IND-03** were not archived in the available benchmark handover. IND-01 and IND-02 point to the same missing pre-Module-1 source; IND-03 points to a separate missing source. Their compiled MuJoCo worlds are included. The index marks these three records as `not_archived`; a different Blender scene is not substituted for them.
 
-## Download and verify
+## Download
 
-Use Python 3.10. From the downloaded dataset root:
+For the lightweight scene index, previews, and portable runtime:
 
 ```bash
-hf download AIGeeksGroup/SimCode4D --repo-type dataset --local-dir SimCode4D
-cd SimCode4D
-python -m pip install -r requirements-benchmark.txt
-python benchmark/simcode4d_unified/check_public_release.py
+hf download AIGeeksGroup/SimCode4D --repo-type dataset \
+  --include "data/*" --include "previews/*" --include "runtime/*" \
+  --local-dir SimCode4D
+tar -xzf SimCode4D/runtime/simcode4d_dev15_runtime.tar.gz -C SimCode4D
 ```
 
-The integrity check verifies the frozen references and loads all 15 compiled
-M1 worlds and runtime models with MuJoCo. The expected output has `"passed":
-true` and `"case_count": 15`.
+The `runtime_model` field in each `data/dev15.jsonl` row points to a MuJoCo XML inside the extracted archive. For example:
+
+```python
+import mujoco
+model = mujoco.MjModel.from_xml_path(
+    "SimCode4D/benchmark/simcode4d_unified/artifacts/INS-01-agv-aisle/runtime_model.xml"
+)
+print(model.ngeom)
+```
 
-To test the submission interface without generating a new world:
+Download a Blender project separately when needed:
 
 ```bash
-python benchmark/simcode4d_unified/tools/make_reference_smoke.py \
-  --output /tmp/simcode4d-reference-package
-python benchmark/simcode4d_unified/tools/score_submission.py \
-  --submission-root /tmp/simcode4d-reference-package \
-  --output /tmp/simcode4d-reference-score \
-  --case-ids SER-03-hotel-drawer --physics-seeds 401
+hf download AIGeeksGroup/SimCode4D blender_sources/INS-01-agv-aisle/scene.blend \
+  --repo-type dataset --local-dir SimCode4D
 ```
 
-This is a reference-interface smoke test, not a generated-method score. See
-the [benchmark README](benchmark/simcode4d_unified/README.md) for the full
-matrix, method submission format, and metric definitions.
-
-## Scope and provenance
-
-The registered reference matrix contains 15 cases × 3 deterministic rebuild
-slots × 5 physics seeds = 225 executions. The three rebuild slots do not
-represent independent generated worlds; there are 75 distinct case-physics
-trials. The reference-probe results described in the benchmark README check
-benchmark self-consistency and must not be presented as a Code2Worlds-2 model
-score or as a held-out leaderboard.
-
-This portable release excludes archived Blender authoring scenes, ablation
-configurations, completed run logs, leaderboards, and demo videos. The
-original authoring sources for three indoor scenes were unavailable. Three
-cases have path-only asset portability repairs; see
-[PORTABILITY.md](benchmark/simcode4d_unified/PORTABILITY.md). The source-name
-binding metric requires authoring-source catalogs outside this compact
-release; the included runtime binding check uses compiled model interfaces.
-
-No repository-wide license was specified in the source release. Third-party
-model assets retain their license and notice files under `vendor/`; review
-those notices before reuse beyond the benchmark runtime.
+Opening authoring projects may require Blender and the original scene-generation dependencies. The MuJoCo runtime archive is the portable path for inspecting and executing the benchmark worlds.
+
+## Evaluation
+
+Use the [GitHub benchmark guide](https://github.com/AIGeeksGroup/Code2Worlds-2/tree/main/benchmark/simcode4d_unified) to run the reference integrity check or score a generated world. It defines WorldBuildPass@1, TaskOverlayPass@1, entity binding F1, task success, mechanics, kinematics, contact/friction, safety, and robust task success. The GitHub release retains the frozen verification metadata needed for reproducible scoring; this data repository presents the scenes and their previews without repeating those metadata tables or shipping a second evaluator.
+
+The split is **development/validation**, not an unseen test set. The reference matrix checks benchmark self-consistency; it is not a Code2Worlds-2 method score.
+
+## Provenance and use
+
+The portable runtime data comes from the public Code2Worlds-2 development suite. The Blender projects are the files named in that suite's source records; all 12 available files were checked against the source records before upload. Third-party model notices are in [`THIRD_PARTY_NOTICES.md`](THIRD_PARTY_NOTICES.md). No repository-wide license is asserted here; review the notices and the applicable project terms before reuse.

THIRD_PARTY_NOTICES.md

@@ -0,0 +1,11 @@
+# Third-party source and assets
+
+The included Code2Worlds source is a snapshot of [`AIGeeksGroup/Code2Worlds`](https://github.com/AIGeeksGroup/Code2Worlds) at commit `03d7edb9242b4095bfb9b3315b1fd67a20d433a1`, with the local integration changes documented in [`third_party/Code2Worlds/UPSTREAM.md`](third_party/Code2Worlds/UPSTREAM.md). The inspected upstream snapshot contains no top-level license file; no license is invented here.
+
+The benchmark retains only the robot meshes needed for its frozen runnable models. License notices are preserved at:
+
+- [`benchmark/simcode4d_unified/vendor/handover9_20260921/unitree_go2/LICENSE`](benchmark/simcode4d_unified/vendor/handover9_20260921/unitree_go2/LICENSE)
+- [`benchmark/simcode4d_unified/vendor/handover9_20260921/universal_robots_ur5e/LICENSE`](benchmark/simcode4d_unified/vendor/handover9_20260921/universal_robots_ur5e/LICENSE)
+- [`benchmark/simcode4d_unified/vendor/jackal_ugv/LICENSE.clearpath`](benchmark/simcode4d_unified/vendor/jackal_ugv/LICENSE.clearpath)
+
+Infinigen is not vendored in this release. Install it separately from [princeton-vl/infinigen](https://github.com/princeton-vl/infinigen) and follow its license.

data/dev15.jsonl

@@ -1,15 +1,15 @@
-{"case_gold_path": "benchmark/simcode4d_unified/cases/IND-01-vacuum-living-room/case_gold.json", "case_id": "IND-01-vacuum-living-room", "case_spec_path": "benchmark/simcode4d_unified/cases/IND-01-vacuum-living-room/case_spec.json", "domain": "indoor", "m1_base_path": "benchmark/simcode4d_unified/vendor/handover9_20260921/cases/indoor-1_vacuum_real/M1/module1/scene_enriched.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/IND-01-vacuum-living-room/physics_profile.json", "prompt": "Drive the vacuum east through the living-room corridor, enter the frozen goal, and remain stopped for 0.5 seconds.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/IND-01-vacuum-living-room/runtime_model.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/IND-01-vacuum-living-room/task_confirm.yaml", "task_family": "obstacle_navigation", "world_id": "ind01-living-room"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/IND-02-go2-living-room/case_gold.json", "case_id": "IND-02-go2-living-room", "case_spec_path": "benchmark/simcode4d_unified/cases/IND-02-go2-living-room/case_spec.json", "domain": "indoor", "m1_base_path": "benchmark/simcode4d_unified/vendor/handover9_20260921/cases/indoor-2_go2lr_walk/M1/module1/scene_enriched.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/IND-02-go2-living-room/physics_profile.json", "prompt": "Traverse the living room east with the Go2, stay in the frozen corridor, enter the goal, and stop.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/IND-02-go2-living-room/runtime_model.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/IND-02-go2-living-room/task_confirm.yaml", "task_family": "obstacle_navigation", "world_id": "ind02-living-room"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/IND-03-vacuum-west/case_gold.json", "case_id": "IND-03-vacuum-west", "case_spec_path": "benchmark/simcode4d_unified/cases/IND-03-vacuum-west/case_spec.json", "domain": "indoor", "m1_base_path": "benchmark/simcode4d_unified/vendor/handover9_20260921/cases/indoor-3_indoor2_v2/M1/module1/scene_enriched.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/IND-03-vacuum-west/physics_profile.json", "prompt": "Drive the vacuum west through the indoor corridor, enter the frozen goal, and remain stopped for 0.5 seconds.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/IND-03-vacuum-west/runtime_model.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/IND-03-vacuum-west/task_confirm.yaml", "task_family": "obstacle_navigation", "world_id": "ind03-indoor-west"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/INS-01-agv-aisle/case_gold.json", "case_id": "INS-01-agv-aisle", "case_spec_path": "benchmark/simcode4d_unified/cases/INS-01-agv-aisle/case_spec.json", "domain": "industrial", "m1_base_path": "benchmark/simcode4d_unified/vendor/handover9_20260921/cases/industrial-1_agv_empty/M1/module1/scene_enriched.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/INS-01-agv-aisle/physics_profile.json", "prompt": "Drive the four-wheel AGV east along the industrial aisle, hold the lane, enter the goal, and stop.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/INS-01-agv-aisle/runtime_model.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/INS-01-agv-aisle/task_confirm.yaml", "task_family": "obstacle_navigation", "world_id": "ins01-industrial-aisle"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/INS-02-agv-cargo/case_gold.json", "case_id": "INS-02-agv-cargo", "case_spec_path": "benchmark/simcode4d_unified/cases/INS-02-agv-cargo/case_spec.json", "domain": "industrial", "m1_base_path": "benchmark/simcode4d_unified/vendor/handover9_20260921/cases/industrial-2_agv_cargo/M1/module1/scene_enriched.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/INS-02-agv-cargo/physics_profile.json", "prompt": "Transport the physically modeled cargo east with the AGV, retain the declared weld attachment, enter the cargo goal, and stop.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/INS-02-agv-cargo/runtime_model.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/INS-02-agv-cargo/task_confirm.yaml", "task_family": "dynamic_interaction", "world_id": "ins02-industrial-cargo"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/INS-03-weld/case_gold.json", "case_id": "INS-03-weld", "case_spec_path": "benchmark/simcode4d_unified/cases/INS-03-weld/case_spec.json", "domain": "industrial", "m1_base_path": "benchmark/simcode4d_unified/vendor/handover9_20260921/cases/industrial-3_weld/M1/module1/scene_enriched.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/INS-03-weld/physics_profile.json", "prompt": "Trace the frozen weld seam continuously with the UR5e tool tip and finish at rest.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/INS-03-weld/runtime_model.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/INS-03-weld/task_confirm.yaml", "task_family": "dynamic_interaction", "world_id": "ins03-welding-cell"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/NAT-01-rock-kick/case_gold.json", "case_id": "NAT-01-rock-kick", "case_spec_path": "benchmark/simcode4d_unified/cases/NAT-01-rock-kick/case_spec.json", "domain": "natural_outdoor", "m1_base_path": "benchmark/simcode4d_unified/artifacts/NAT-01-rock-kick/m1_world.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/NAT-01-rock-kick/physics_profile.json", "prompt": "Use the Go2 feet to make the source-bound rock move physically in the uneven canyon, then let it settle.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/NAT-01-rock-kick/m1_world.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/NAT-01-rock-kick/task_confirm.yaml", "task_family": "dynamic_interaction", "world_id": "nat01-canyon"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/NAT-02-hills-turn-stop/case_gold.json", "case_id": "NAT-02-hills-turn-stop", "case_spec_path": "benchmark/simcode4d_unified/cases/NAT-02-hills-turn-stop/case_spec.json", "domain": "natural_outdoor", "m1_base_path": "benchmark/simcode4d_unified/artifacts/NAT-02-hills-turn-stop/m1_world.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/NAT-02-hills-turn-stop/physics_profile.json", "prompt": "Drive the Jackal across the authored hills, complete the authored left turn, enter the goal region, and remain stopped for 0.5 seconds.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/NAT-02-hills-turn-stop/m1_world.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/NAT-02-hills-turn-stop/task_confirm.yaml", "task_family": "obstacle_navigation", "world_id": "nat02-hills"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/NAT-03-water-land/case_gold.json", "case_id": "NAT-03-water-land", "case_spec_path": "benchmark/simcode4d_unified/cases/NAT-03-water-land/case_spec.json", "domain": "natural_outdoor", "m1_base_path": "benchmark/simcode4d_unified/artifacts/NAT-03-water-land/m1_world.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/NAT-03-water-land/physics_profile.json", "prompt": "Move the amphibious rover from the declared static water field, contact the bank, reach dry land, and stop in the land goal.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/NAT-03-water-land/m1_world.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/NAT-03-water-land/task_confirm.yaml", "task_family": "dynamic_interaction", "world_id": "nat03-amphibious"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/SER-01-medicine/case_gold.json", "case_id": "SER-01-medicine", "case_spec_path": "benchmark/simcode4d_unified/cases/SER-01-medicine/case_spec.json", "domain": "service", "m1_base_path": "benchmark/simcode4d_unified/artifacts/SER-01-medicine/m1_world.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/SER-01-medicine/physics_profile.json", "prompt": "Approach, legally grasp, and deliver the medicine pouch to the authored delivery region while retaining a stable grasp.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/SER-01-medicine/m1_world.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/SER-01-medicine/task_confirm.yaml", "task_family": "pick_and_place", "world_id": "ser01-clinic"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/SER-02-airport-luggage/case_gold.json", "case_id": "SER-02-airport-luggage", "case_spec_path": "benchmark/simcode4d_unified/cases/SER-02-airport-luggage/case_spec.json", "domain": "service", "m1_base_path": "benchmark/simcode4d_unified/artifacts/SER-02-airport-luggage/m1_world.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/SER-02-airport-luggage/physics_profile.json", "prompt": "Approach the blue luggage assembly, legally grasp and lift it, then deliver it to the authored airport goal while retaining a stable grasp.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/SER-02-airport-luggage/m1_world.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/SER-02-airport-luggage/task_confirm.yaml", "task_family": "pick_and_place", "world_id": "ser02-airport"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/SER-03-hotel-drawer/case_gold.json", "case_id": "SER-03-hotel-drawer", "case_spec_path": "benchmark/simcode4d_unified/cases/SER-03-hotel-drawer/case_spec.json", "domain": "service", "m1_base_path": "benchmark/simcode4d_unified/artifacts/SER-03-hotel-drawer/m1_world.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/SER-03-hotel-drawer/physics_profile.json", "prompt": "Navigate the hotel robot to the terminal region, stop, open its payload drawer, and keep the meal parcel supported for 0.5 seconds.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/SER-03-hotel-drawer/m1_world.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/SER-03-hotel-drawer/task_confirm.yaml", "task_family": "dynamic_interaction", "world_id": "ser03-hotel"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/URB-01-agv-city/case_gold.json", "case_id": "URB-01-agv-city", "case_spec_path": "benchmark/simcode4d_unified/cases/URB-01-agv-city/case_spec.json", "domain": "urban", "m1_base_path": "benchmark/simcode4d_unified/vendor/handover9_20260921/cases/urban-1_agv_city/M1/module1/scene_enriched.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/URB-01-agv-city/physics_profile.json", "prompt": "Drive the AGV east along the city pavement, stay in the frozen corridor, enter the goal, and stop.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/URB-01-agv-city/runtime_model.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/URB-01-agv-city/task_confirm.yaml", "task_family": "obstacle_navigation", "world_id": "urb01-city-pavement"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/URB-02-intersection/case_gold.json", "case_id": "URB-02-intersection", "case_spec_path": "benchmark/simcode4d_unified/cases/URB-02-intersection/case_spec.json", "domain": "urban", "m1_base_path": "benchmark/simcode4d_unified/vendor/handover9_20260921/cases/urban-2_intersection/M1/module1/scene_enriched.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/URB-02-intersection/physics_profile.json", "prompt": "Start on the west sidewalk, cross the road east through the frozen corridor, reach the east sidewalk, and stop.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/URB-02-intersection/runtime_model.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/URB-02-intersection/task_confirm.yaml", "task_family": "obstacle_navigation", "world_id": "urb02-intersection"}
-{"case_gold_path": "benchmark/simcode4d_unified/cases/URB-03-pole-blocking/case_gold.json", "case_id": "URB-03-pole-blocking", "case_spec_path": "benchmark/simcode4d_unified/cases/URB-03-pole-blocking/case_spec.json", "domain": "urban", "m1_base_path": "benchmark/simcode4d_unified/vendor/handover9_20260921/cases/urban-3_delivery/M1/module1/scene_enriched.xml", "physics_profile_path": "benchmark/simcode4d_unified/cases/URB-03-pole-blocking/physics_profile.json", "prompt": "Approach the frozen target pole with the Go2, make physical contact, remain blocked without passing through it, and stop.", "runtime_model_path": "benchmark/simcode4d_unified/artifacts/URB-03-pole-blocking/runtime_model.xml", "split": "development", "task_confirm_path": "benchmark/simcode4d_unified/cases/URB-03-pole-blocking/task_confirm.yaml", "task_family": "dynamic_interaction", "world_id": "urb03-pole-blocking"}
+{"case_id": "IND-01-vacuum-living-room", "domain": "Indoor", "task_family": "obstacle_navigation", "dynamic_subtype": null, "prompt": "Drive the vacuum east through the living-room corridor, enter the frozen goal, and remain stopped for 0.5 seconds.", "runtime_model": "benchmark/simcode4d_unified/artifacts/IND-01-vacuum-living-room/runtime_model.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": null, "source_status": "not_archived", "preview_kind": "MuJoCo scene views", "preview_1": "previews/IND-01-vacuum-living-room/01_overview.png", "preview_2": "previews/IND-01-vacuum-living-room/02_agent.png", "preview_3": "previews/IND-01-vacuum-living-room/03_task.png"}
+{"case_id": "IND-02-go2-living-room", "domain": "Indoor", "task_family": "obstacle_navigation", "dynamic_subtype": null, "prompt": "Traverse the living room east with the Go2, stay in the frozen corridor, enter the goal, and stop.", "runtime_model": "benchmark/simcode4d_unified/artifacts/IND-02-go2-living-room/runtime_model.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": null, "source_status": "not_archived", "preview_kind": "MuJoCo scene views", "preview_1": "previews/IND-02-go2-living-room/01_overview.png", "preview_2": "previews/IND-02-go2-living-room/02_agent.png", "preview_3": "previews/IND-02-go2-living-room/03_task.png"}
+{"case_id": "IND-03-vacuum-west", "domain": "Indoor", "task_family": "obstacle_navigation", "dynamic_subtype": null, "prompt": "Drive the vacuum west through the indoor corridor, enter the frozen goal, and remain stopped for 0.5 seconds.", "runtime_model": "benchmark/simcode4d_unified/artifacts/IND-03-vacuum-west/runtime_model.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": null, "source_status": "not_archived", "preview_kind": "MuJoCo scene views", "preview_1": "previews/IND-03-vacuum-west/01_overview.png", "preview_2": "previews/IND-03-vacuum-west/02_agent.png", "preview_3": "previews/IND-03-vacuum-west/03_task.png"}
+{"case_id": "INS-01-agv-aisle", "domain": "Industrial", "task_family": "obstacle_navigation", "dynamic_subtype": null, "prompt": "Drive the four-wheel AGV east along the industrial aisle, hold the lane, enter the goal, and stop.", "runtime_model": "benchmark/simcode4d_unified/artifacts/INS-01-agv-aisle/runtime_model.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": "blender_sources/INS-01-agv-aisle/scene.blend", "source_status": "archived", "preview_kind": "MuJoCo scene views", "preview_1": "previews/INS-01-agv-aisle/01_overview.png", "preview_2": "previews/INS-01-agv-aisle/02_agent.png", "preview_3": "previews/INS-01-agv-aisle/03_task.png"}
+{"case_id": "INS-02-agv-cargo", "domain": "Industrial", "task_family": "dynamic_interaction", "dynamic_subtype": "contact_dynamic", "prompt": "Transport the physically modeled cargo east with the AGV, retain the declared weld attachment, enter the cargo goal, and stop.", "runtime_model": "benchmark/simcode4d_unified/artifacts/INS-02-agv-cargo/runtime_model.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": "blender_sources/INS-02-agv-cargo/scene.blend", "source_status": "archived", "preview_kind": "MuJoCo scene views", "preview_1": "previews/INS-02-agv-cargo/01_overview.png", "preview_2": "previews/INS-02-agv-cargo/02_agent.png", "preview_3": "previews/INS-02-agv-cargo/03_task.png"}
+{"case_id": "INS-03-weld", "domain": "Industrial", "task_family": "dynamic_interaction", "dynamic_subtype": "articulated", "prompt": "Trace the frozen weld seam continuously with the UR5e tool tip and finish at rest.", "runtime_model": "benchmark/simcode4d_unified/artifacts/INS-03-weld/runtime_model.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": "blender_sources/INS-03-weld/scene.blend", "source_status": "archived", "preview_kind": "MuJoCo scene views", "preview_1": "previews/INS-03-weld/01_overview.png", "preview_2": "previews/INS-03-weld/02_agent.png", "preview_3": "previews/INS-03-weld/03_task.png"}
+{"case_id": "NAT-01-rock-kick", "domain": "Natural outdoor", "task_family": "dynamic_interaction", "dynamic_subtype": "contact_dynamic", "prompt": "Use the Go2 feet to make the source-bound rock move physically in the uneven canyon, then let it settle.", "runtime_model": "benchmark/simcode4d_unified/artifacts/NAT-01-rock-kick/m1_world.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": "blender_sources/NAT-01-rock-kick/scene.blend", "source_status": "archived", "preview_kind": "Blender animation frames", "preview_1": "previews/NAT-01-rock-kick/01_start.png", "preview_2": "previews/NAT-01-rock-kick/02_middle.png", "preview_3": "previews/NAT-01-rock-kick/03_end.png"}
+{"case_id": "NAT-02-hills-turn-stop", "domain": "Natural outdoor", "task_family": "obstacle_navigation", "dynamic_subtype": null, "prompt": "Drive the Jackal across the authored hills, complete the authored left turn, enter the goal region, and remain stopped for 0.5 seconds.", "runtime_model": "benchmark/simcode4d_unified/artifacts/NAT-02-hills-turn-stop/m1_world.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": "blender_sources/NAT-02-hills-turn-stop/scene.blend", "source_status": "archived", "preview_kind": "Blender animation frames", "preview_1": "previews/NAT-02-hills-turn-stop/01_start.png", "preview_2": "previews/NAT-02-hills-turn-stop/02_middle.png", "preview_3": "previews/NAT-02-hills-turn-stop/03_end.png"}
+{"case_id": "NAT-03-water-land", "domain": "Natural outdoor", "task_family": "dynamic_interaction", "dynamic_subtype": "cross_medium", "prompt": "Move the amphibious rover from the declared static water field, contact the bank, reach dry land, and stop in the land goal.", "runtime_model": "benchmark/simcode4d_unified/artifacts/NAT-03-water-land/m1_world.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": "blender_sources/NAT-03-water-land/scene.blend", "source_status": "archived", "preview_kind": "Blender animation frames", "preview_1": "previews/NAT-03-water-land/01_start.png", "preview_2": "previews/NAT-03-water-land/02_middle.png", "preview_3": "previews/NAT-03-water-land/03_end.png"}
+{"case_id": "SER-01-medicine", "domain": "Service", "task_family": "pick_and_place", "dynamic_subtype": null, "prompt": "Approach, legally grasp, and deliver the medicine pouch to the authored delivery region while retaining a stable grasp.", "runtime_model": "benchmark/simcode4d_unified/artifacts/SER-01-medicine/m1_world.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": "blender_sources/SER-01-medicine/scene.blend", "source_status": "archived", "preview_kind": "Blender animation frames", "preview_1": "previews/SER-01-medicine/01_start.png", "preview_2": "previews/SER-01-medicine/02_middle.png", "preview_3": "previews/SER-01-medicine/03_end.png"}
+{"case_id": "SER-02-airport-luggage", "domain": "Service", "task_family": "pick_and_place", "dynamic_subtype": null, "prompt": "Approach the blue luggage assembly, legally grasp and lift it, then deliver it to the authored airport goal while retaining a stable grasp.", "runtime_model": "benchmark/simcode4d_unified/artifacts/SER-02-airport-luggage/m1_world.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": "blender_sources/SER-02-airport-luggage/scene.blend", "source_status": "archived", "preview_kind": "Blender animation frames", "preview_1": "previews/SER-02-airport-luggage/01_start.png", "preview_2": "previews/SER-02-airport-luggage/02_middle.png", "preview_3": "previews/SER-02-airport-luggage/03_end.png"}
+{"case_id": "SER-03-hotel-drawer", "domain": "Service", "task_family": "dynamic_interaction", "dynamic_subtype": "articulated", "prompt": "Navigate the hotel robot to the terminal region, stop, open its payload drawer, and keep the meal parcel supported for 0.5 seconds.", "runtime_model": "benchmark/simcode4d_unified/artifacts/SER-03-hotel-drawer/m1_world.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": "blender_sources/SER-03-hotel-drawer/scene.blend", "source_status": "archived", "preview_kind": "Blender animation frames", "preview_1": "previews/SER-03-hotel-drawer/01_start.png", "preview_2": "previews/SER-03-hotel-drawer/02_middle.png", "preview_3": "previews/SER-03-hotel-drawer/03_end.png"}
+{"case_id": "URB-01-agv-city", "domain": "Urban", "task_family": "obstacle_navigation", "dynamic_subtype": null, "prompt": "Drive the AGV east along the city pavement, stay in the frozen corridor, enter the goal, and stop.", "runtime_model": "benchmark/simcode4d_unified/artifacts/URB-01-agv-city/runtime_model.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": "blender_sources/URB-01-agv-city/scene.blend", "source_status": "archived", "preview_kind": "MuJoCo scene views", "preview_1": "previews/URB-01-agv-city/01_overview.png", "preview_2": "previews/URB-01-agv-city/02_agent.png", "preview_3": "previews/URB-01-agv-city/03_task.png"}
+{"case_id": "URB-02-intersection", "domain": "Urban", "task_family": "obstacle_navigation", "dynamic_subtype": null, "prompt": "Start on the west sidewalk, cross the road east through the frozen corridor, reach the east sidewalk, and stop.", "runtime_model": "benchmark/simcode4d_unified/artifacts/URB-02-intersection/runtime_model.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": "blender_sources/URB-02-intersection/scene.blend", "source_status": "archived", "preview_kind": "MuJoCo scene views", "preview_1": "previews/URB-02-intersection/01_overview.png", "preview_2": "previews/URB-02-intersection/02_agent.png", "preview_3": "previews/URB-02-intersection/03_task.png"}
+{"case_id": "URB-03-pole-blocking", "domain": "Urban", "task_family": "dynamic_interaction", "dynamic_subtype": "contact_dynamic", "prompt": "Approach the frozen target pole with the Go2, make physical contact, remain blocked without passing through it, and stop.", "runtime_model": "benchmark/simcode4d_unified/artifacts/URB-03-pole-blocking/runtime_model.xml", "runtime_archive": "runtime/simcode4d_dev15_runtime.tar.gz", "blend_file": "blender_sources/URB-03-pole-blocking/scene.blend", "source_status": "archived", "preview_kind": "MuJoCo scene views", "preview_1": "previews/URB-03-pole-blocking/01_overview.png", "preview_2": "previews/URB-03-pole-blocking/02_agent.png", "preview_3": "previews/URB-03-pole-blocking/03_task.png"}