README.md

---

pretty_name: RecoverBench
license: mit
task_categories:
  - robotics
tags:
  - robot-manipulation
  - error-recovery
  - mujoco
  - robosuite
---


# RecoverBench

Version **1.1.1**. Data, images, indexes and documentation are distributed together in the standard archives.

RecoverBench is a systematic benchmark for **error recovery in robotic manipulation**:
a taxonomy of 12 Error Skills × 2 difficulty degrees (24 subtypes, grouped into 5
Recovery Behavior Groups A–E), an automatic pipeline that injects errors into clean
trajectories and snapshots reproducible post-error states, and a recovery-demonstration
corpus for training recovery-capable policies. Tasks: `pick_place`, `stack`, `coffee`,
`threading`, `stack_three`, `three_piece_assembly` (robosuite / MuJoCo; mixed
Sawyer + Panda embodiments).

Code, evaluation protocol, and regeneration pipeline:
https://github.com/ErrorRecoveryBenchmark/ErrorRecoveryBenchmark

## Files

| File | Contents |
|---|---|
| `eval_data.tar` | **Evaluation set**: 1,360 error scenes (136 task × error-subtype combinations × 10 scenes). Each scene = `<id>.npz` (full MuJoCo state snapshot `post_sim_state`, quaternions in `(w,x,y,z)` order) + `<id>.json` (labels, environment fingerprint, RNG state). Layout: `error_scenes/<task>/scenes/`. |
| `training_scenes.tar` | **Training scene pool**: 5,038 error scenes generated by the same pipeline, disjoint from the evaluation set (verified: zero shared scene IDs and zero identical `post_sim_state` fingerprints). Recovery demonstrations were collected on scenes from this pool. Layout: `training_scenes/<task>/scenes/`. |
| `training_data.tar` | **Recovery demonstrations**: 973 human SpaceMouse teleoperation entries (`recovery_demos_human/`) and MimicGen-style augmented demos (`recovery_demos_augmented/`, 8,957 trajectories). Each demo = `.npz` with `actions`, `states`, `eef_positions`, `target_poses`, `gripper_states`. |
| `recovery_mimicgen_generated.tar` | 668 generated recovery trajectories; part of the 10,598-entry corpus. |
| `recovery_training_success_from_error.tar` | Separate 739-trajectory HDF5 training artifact for the reported recovery-policy experiments. |
| `SHA256SUMS` | SHA-256 checksums for the five data archives. |
| `eval_scene_ids.tsv` | The frozen evaluation-scene ID list (task ↔ scene id). |

## Quick start

```bash

# from the code repository

bash scripts/download_data.sh /path/to/data

export BENCHMARK_DATA=/path/to/data

```

Evaluation protocol (success = 10 consecutive success frames, 500-step budget,
single seed per scene) and metrics are specified in the code repository's
`EVALUATION.md`.

## Reproducibility notes

- Simulator stack is version-locked: robosuite 1.4.1 @ `c848ca84`, `mujoco==2.3.2`,
  `numpy<2.0` (see `ROBOSUITE_VERSION_LOCK.txt` in the code repo).
- Every error scene carries an environment fingerprint and RNG state for
  deterministic restoration via the code repo's replay tooling.
- Train/eval disjointness was verified at the level of scene IDs and SHA-256
  fingerprints of the canonical post-error simulator state.

## Citation

Zijia Tang, Ganlong Zhao, Xingping Chen, Junye Chen, Zihao Mo, Zedi Wang, Guanbin Li. RecoverBench: A Systematic Benchmark for Error Recovery in Robotic Manipulation. NeurIPS 2026, Evaluations and Datasets Track (accepted).

## License and provenance

The RecoverBench release is licensed under MIT; see LICENSE. Upstream simulator,
MimicGen source code, and seed datasets retain their own terms and attribution.
Seeds: https://huggingface.co/datasets/amandlek/mimicgen_datasets.

MimicGen: https://github.com/NVlabs/mimicgen (NVIDIA Source Code License).



Created and maintained by Zijia Tang, Ganlong Zhao, Xingping Chen, Junye Chen, Zihao Mo, Zedi Wang, Guanbin Li. Affiliations are in `CITATION.cff`. Correspondence: liguanbin@mail.sysu.edu.cn; zt81@duke.edu. Funding: NSFC 62322608 and Shenzhen Loop Area Institute FPF10120260001; compute: National Supercomputer Center in Guangzhou. The authors declare no competing interests; the submitting author confirmed all-coauthor agreement on October 21, 2026.

Use the code repository issue tracker for corrections.



## Machine-readable records



`eval_scenes.csv`, `training_scenes.csv`, and `recovery_demos.csv` index every NPZ
member of the core archives. `croissant.json` provides typed CSV column mappings,
archive checksums, RAI metadata, and upstream provenance. NPZ numerical arrays are
loaded separately with NumPy; the CSV indexes do not imply tensor decoding support.
The inventory verifies zero shared scene IDs and exact post-state fingerprints
between the frozen evaluation and training scene archives.

## Limitations and intended use

Simulation-only tabletop recovery evaluation and policy training; mixed Sawyer and
Panda embodiments, six tasks, hand-designed failure types, and uneven subtype
coverage. This is not evidence of safe real-world deployment or general-purpose
robot capability. Report clean-state and recovery success together.

The corpus contains 973 human, 668 generated, and 8,957 augmented trajectories (10,598 total). The separate 739 HDF5 trajectories are the recovery-policy training artifact, not additional corpus members.

## Data format and reproducibility

Images are stored in `camera_images` inside their demonstration NPZ. Where present,
`camera_state_indices` and `camera_state_times` specify alignment to saved states;
`camera_source` and `camera_metadata` describe the rendering process. Image counts
and sampling rates vary between records. See [DATA_FORMAT.md](DATA_FORMAT.md).

The corpus has 973 human, 668 generated, and 8,957 augmented trajectories (10,598
total). The 739 HDF5 training trajectories are listed separately. The HDF5 files
retain historical model asset paths; normalize these with the pinned environment's
`edit_model_xml` before replay. Structural checks do not certify action replay.
Five CSV indexes expose the archive records to Croissant. Version history and
record-level provenance are retained for reproducibility.