Skip to content

AstroAgentBench v2.0.0

Latest
Latest

Choose a tag to compare

View all tags
@Mtrya Mtrya released this 17 Oct 15:24
· 5 commits to main since this release
v2.0.0
6f5b9d2
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

AstroAgentBench v2.0.0 publishes the benchmark in Harbor-native formats for reproducible agent evaluation.

  • RL Environment taskset: 131 tasks across seven benchmark configurations, with native Harbor loading and dataset viewing.
  • Agent traces: 115 Claude Code and Codex sessions rendered by Hugging Face's native trace viewer. The source archive is preserved on Zenodo.
  • Runtime image: ghcr.io/mtrya/astroagentbench-python@sha256:fe9733cad55519aa08dd0db9abb51444051ef5f53d6ac1ef0b4b6592491437d1.

All seven oracle smoke checks passed against the promoted Hugging Face taskset. Publication evidence: taskset workflow and traces workflow.

Assets 2
Loading