Pinned Loading

1. jlens-rl jlens-rl Public

   Jacobian-lens monitoring of reward hacking under RL

   Python

2. probex probex Public

   Inline safety probes on production LLM servers

   Python

3. cua-bench cua-bench Public

   Computer-use agent benchmark on ERP accounting

   Python

4. soma-org/soma soma-org/soma Public

   Open Collective Superintelligence

   Rust 10

5. probe-necessity probe-necessity Public

   Necessity tests for linear safety probes

   Python

6. monitor-drift monitor-drift Public

   How fine-tuning breaks probe-based safety monitors

   Python