{"data":{"kind":"file","path":"README.md","version_id":"igd8wfi7yphgl1iteeymrczu","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":602,"modified_at":"2026-09-16T12:25:28.037000","content_hash":"3a27bea94809990dc68547e41fa7c389afc5659a9317a35c7d4acfdd9f5e0462"},"entries":[],"content":"# mmmu\n\nMMMU multimodal multiple-choice questions spanning university-level subjects. Each task interleaves the question text with its referenced images and scores the final boxed option letter.\n\n## Taskset\n\n- **Source:** [`MMMU/MMMU`](https://huggingface.co/datasets/MMMU/MMMU)\n- **Splits:** `dev` (default), `validation`, and `test`\n- **Subsets:** 30 subject areas, or one selected with `--env.taskset.subset`\n- **Scoring:** Exact match on the option letter inside `\\boxed{}`; open-ended rows are skipped\n\n## Changelog\n\n- 2026-07-16: Moved the V1 taskset from the Verifiers examples into prime-envs.\n","encoding":"utf-8","truncated":false,"total_bytes":602},"status":null}