- Add Harbor tasks using the published ExtractBench coding-agent prompt (32a2e81c091212a29095eaa9f2d514daa047cb3c)
- Document Git LFS prerequisite for Harbor (da58a6e50cffe7c89f618fef32366e1c3efe77d6)
- Add aligned cell feedback for the existing Value F1 reward (6ca3001dc7d9ab23568ac2e4f5939e2c82978828)
- Set the agent timeout to 1800 seconds for every task (7f605a63ed0381185ae29bc32b1493d6a4fad670)
- Enforce agent-phase model-host allowlists and offline verification (3d8569400fbb04bdeffde38e12faab14bf4ad428)
- Remove RL Environment classification (d7962b8926d402a05a35eb60c74ac8d60d7d84ee)
Test RL Environment badge precedence on registered ExtractBench
Adds only the `rl-environment` dataset-card tag to test its display alongside the existing official Benchmark registration. No Harbor export files are included.
The existing `eval.yaml`, dataset configuration, documents, and annotations are unchanged. Badge precedence remains to be checked after merge.
29 days ago
ExtractBench v1.1: tighter boxes, new box ground truth, value fixes (#2)
