Pinned Loading

1. CurationGym CurationGym Public

   compute-budget-aware curation optimizer that searches filtering + mixing policies

   Python

2. label-forge label-forge Public

   Deterministic pipeline for multimodal labeling and synthetic data generation

   Python

3. mmeval-lab mmeval-lab Public

   Unified multimodal regression harness that runs MMMU + OmniDocBench + Video-MME, and adds contamination scans, slice mining, and data-diff attribution to explain which dataset changes drive which b…

   Python 1

4. embodied-agent embodied-agent Public

   A dataset compiler that converts heterogeneous robot trajectories (RLDS, LeRobot, Open X-Embodiment) into a streamable storage format with strict validation and sliceable metadata.

   Python 1