# ManiSkill (/docs/evaluation/benchmark-hub/maniskill)



## About [#about]

ManiSkill external benchmark task manifest for simulator-backed manipulation and planning rollouts.

## Metrics [#metrics]

ManiSkill external benchmark task manifest for simulator-backed manipulation and planning rollouts.

### Task YAML [#task-yaml]

`worldfoundry/data/benchmarks/tasks/external/maniskill.yaml`

### Primary [#primary]

* `success_rate`

| Metric              | Focus                                                                                                          |
| ------------------- | -------------------------------------------------------------------------------------------------------------- |
| `success_rate`      | Share of episodes or tasks that complete successfully.                                                         |
| `task_success`      | Share of episodes or tasks that complete successfully.                                                         |
| `reward`            | Undiscounted task reward accumulated over official ManiSkill SAPIEN manipulation rollouts.                     |
| `normalized_return` | ManiSkill episode return scaled against a reference expert or random-policy ceiling for cross-task comparison. |

## Environment setup [#environment-setup]

### Runtime environment [#runtime-environment]

* **Conda env:** `maniskill`.
* **Separate setup:** defaults to the unified WorldFoundry env (`needs_new_env: false`); a benchmark-only conda env is usually not required.
* ManiSkill/SAPIEN 可见；完整官方状态仍需要任务套件、assets 和 policy rollout，不在本轮运行。

### Evaluation data [#evaluation-data]

* ManiSkill is simulator-backed; WorldFoundry normalizes rollout results and keeps simulator execution explicit.
* Candidate model outputs: set `WORLDFOUNDRY_MANISKILL_RESULTS_PATH` to the generated-video or rollout artifact root.
* `${WORLDFOUNDRY_CACHE_DIR}/repos/haosulab--ManiSkill`

### Checkpoints & assets [#checkpoints--assets]

* This benchmark is primarily rollout- or official-result-driven; see catalog `requires` and the task YAML for policy/simulator checkpoints.

### Key environment variables [#key-environment-variables]

* `WORLDFOUNDRY_MANISKILL_ROOT`
* `WORLDFOUNDRY_MANISKILL_RESULTS_PATH`

### Verify setup [#verify-setup]

```bash
PYTHONPATH=${WORLDFOUNDRY_CACHE_DIR}/repos/haosulab--ManiSkill:src python -m worldfoundry.evaluation.tasks.execution.orchestration.runtime_preflight --profile maniskill --manifest worldfoundry/data/benchmarks/runtime_profiles/official/maniskill.yaml --output-dir tmp/worldfoundry_preflight/maniskill --json
```

## Run Evaluation [#run-evaluation]

After preparing the assets and candidate artifacts listed on this page, use the public run entry points below.

### Set Candidate Artifacts [#set-candidate-artifacts]

```bash
export WORLDFOUNDRY_GENERATED_ARTIFACT_DIR=/path/to/generated/artifacts
```

### Import Official Results [#import-official-results]

```bash
worldfoundry-eval zoo benchmark-run \
  --benchmark-id maniskill \
  --mode normalizer \
  --official-results-path /path/to/official/results-or-report \
  --generated-artifact-dir "${WORLDFOUNDRY_GENERATED_ARTIFACT_DIR}" \
  --output-dir tmp/maniskill/normalizer \
  --json
```

### Score Generated Artifacts [#score-generated-artifacts]

```bash
worldfoundry-eval zoo benchmark-run \
  --benchmark-id maniskill \
  --mode official-run \
  --generated-artifact-dir "${WORLDFOUNDRY_GENERATED_ARTIFACT_DIR}" \
  --output-dir tmp/maniskill/official-run \
  --json
```

### Direct In-Tree Runner [#direct-in-tree-runner]

```bash
worldfoundry-eval zoo benchmark-run --benchmark-id maniskill --mode official-validation --official-results-path <official_results.json> --output-dir <out> --json
```

## Requirements [#requirements]

### Inputs [#inputs]

* Official ManiSkill repository checkout
* SAPIEN/ManiSkill simulator assets
* Benchmark-specific policy checkpoint or controller
* Official task suite selection

### Environment [#environment]

* `WORLDFOUNDRY_MANISKILL_RESULTS_PATH`

### Outputs [#outputs]

* `scorecard.json`
* `raw_results.jsonl`

[← Benchmark Hub](/docs/evaluation/benchmark-hub)