Activity and trending
Stars, forks and merged pull requests each day or month, lined up with the repository's rank on Trendshift Trending
Open-source framework for auditing reward hacking in AI agents
Stars, forks and merged pull requests each day or month, lined up with the repository's rank on Trendshift Trending
4/5 The project explores: • Adversarial scenarios and payload transformations • Per-attempt canaries • Observable-behavior oracles • Replay and structured reports • Statistical summaries and CI…
@subh10 · x.com
2/5 𝐓𝐡𝐞 𝐢𝐝𝐞𝐚 💡: a useful adversarial finding should be inspectable. What scenario ran? What behavior was observed? What evidence supports the result? Can someone replay it? #RL_Hacking_Audit…
@subh10 · x.com