LIBERO-Industry
Harbor: pendingA robotic manipulation benchmark with 216 tasks evaluating generalist policies under industrial constraints like moving flows and tight tolerances.
- Samples
- 216
- Released
- 2026
- Scores
- 0
LIBERO-RIG
LIBERO-RIG: A Hardware-Setup Robustness Benchmark for Vision-Language-Action Models
A robustness benchmark evaluating VLA models across 210 hardware perturbation conditions (11 axes, 5 levels) on the LIBERO suite, measuring success rate degradation under physical rig variations.
How to run
ltb · libero-rig
$ npm i -g lettokenburn-cli$ ltb bench info libero-rig
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
0 sourced entriesNo sourced scores recorded yet.
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsLoading community runs…
About
LIBERO-RIG perturbs robot hardware setups (cameras, calibration, control rates) while keeping tasks fixed, creating 210 conditions across four LIBERO suites. It evaluates models using 500 episodes per condition to assess robustness.
Scoring uses raw success rates averaged over conditions. Findings show significant performance drops at high severity, with latent faults like control frequency mismatches causing distinct failure patterns compared to visual perturbations.