| pretty_name: YALA Robotics Evaluation Artifacts | |
| tags: | |
| - robotics | |
| - simulation | |
| - evaluation | |
| # YALA Robotics Evaluation Artifacts | |
| Public archive of generated evaluation and diagnostic artifacts from YALA manipulation-policy experiments. The archive contains episode records, runtime traces, reset/reproducibility diagnostics, seed manifests, and experiment result files. Source code, training datasets, and policy checkpoints are kept in the YALA repository rather than this dataset. | |
| The files retain their original repository-relative paths under `experiments/` and `results/`. `MANIFEST.csv` records each source path, byte size, and SHA256. | |
| For the fresh `manipulate_pipette` ACT versus FailSafe checkpoint comparison, the two arms use the same numeric seed list. Simulator state restoration was not verified, so same-seed outcome transitions are descriptive only and are not causal recovery counts. | |
| This repository is public and contains simulator-generated data. | |