# LIBERO (/docs/evaluation/benchmark-hub/libero)



## About [#about]

LIBERO is a lifelong robot manipulation benchmark with language-conditioned tasks across spatial, object, goal, and long-horizon suites.

## Metrics [#metrics]

LIBERO external benchmark task manifest for language-conditioned robot manipulation policy rollouts.

### Task YAML [#task-yaml]

`worldfoundry/data/benchmarks/tasks/external/libero.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.                                         |
| `episode_success` | Share of episodes or tasks that complete successfully.                                         |
| `action_accuracy` | Correctness of predicted actions, states, or outcomes against the benchmark success criterion. |

## Environment setup [#environment-setup]

### Runtime environment [#runtime-environment]

* **Conda env:** `giga-world-policy-py311`.
* **Separate setup:** defaults to the unified WorldFoundry env (`needs_new_env: false`); a benchmark-only conda env is usually not required.
* LIBERO Python 包可见；完整状态还需要 policy checkpoint、suite 选择和官方 rollout 结果。

### Evaluation data [#evaluation-data]

* Candidate model outputs: set `WORLDFOUNDRY_LIBERO_RESULTS_PATH` to the generated-video or rollout artifact root.
* `${WORLDFOUNDRY_HFD_DATASET_ROOT}/yifengzhu-hf__LIBERO-datasets`
* `${WORLDFOUNDRY_HFD_DATASET_ROOT}/lerobot__libero`
* `${WORLDFOUNDRY_CACHE_DIR}/repos/Lifelong-Robot-Learning--LIBERO`

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

* Base-model / metric dependencies: `libero_source_assets`, `libero_dataset_assets`.
* Default metric-checkpoint paths and override env vars are listed in the [local assets guide](/docs/guides/local-assets).

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

* `WORLDFOUNDRY_LIBERO_ROOT`
* `WORLDFOUNDRY_LIBERO_RESULTS_PATH`

### Verify setup [#verify-setup]

```bash
PYTHONPATH=${WORLDFOUNDRY_CACHE_DIR}/repos/Lifelong-Robot-Learning--LIBERO:src ${WORLDFOUNDRY_CONDA_ENVS_ROOT}/worldfoundry-giga-world-policy-py311/bin/python -m worldfoundry.evaluation.tasks.execution.orchestration.runtime_preflight --profile libero --manifest worldfoundry/data/benchmarks/runtime_profiles/official/libero.yaml --output-dir tmp/worldfoundry_preflight/libero --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 libero \
  --mode normalizer \
  --official-results-path /path/to/official/results-or-report \
  --generated-artifact-dir "${WORLDFOUNDRY_GENERATED_ARTIFACT_DIR}" \
  --output-dir tmp/libero/normalizer \
  --json
```

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

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

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

```bash
worldfoundry-eval embodied run --config worldfoundry/data/benchmarks/eval_configs/embodied/libero/spatial.yaml
```

## Requirements [#requirements]

### Inputs [#inputs]

* Upstream SAPIEN/Curobo simulator runtime
* RoboTwin simulator assets
* Policy checkpoint for selected upstream eval script
* Official eval\_result tree or structured result export

### Environment [#environment]

* `WORLDFOUNDRY_LIBERO_RESULTS_PATH`

### Outputs [#outputs]

* `scorecard.json`
* `run_manifest.json`
* `results.jsonl`
* `metrics/summary.json`

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