COMPASS

Popular repositories Loading

1. prompt_injections_so_back prompt_injections_so_back Public

   Code and data used in our NeurIPS 2026 paper: AI Agents May Always Fall for Prompt Injections

   Python 13 2

2. SCARCE-ENV SCARCE-ENV Public

   Emergent institutions in long-horizon multi-agent coding systems

   Python 4

3. neurips2026_evaluation_meta_knowledge neurips2026_evaluation_meta_knowledge Public

   Repository for the NeurIPS 2026 E&D "Models That Know How Evaluations Are Designed Score Safer"

   Jupyter Notebook 3 1

4. FIGSBench FIGSBench Public

   FIGS: Evaluating Multi-Turn Sycophancy Without Penalizing Empathy

   HTML 2

5. CORSA-Cluster-Optimization-for-Router-Aware-Skill-Attacks CORSA-Cluster-Optimization-for-Router-Aware-Skill-Attacks Public

   Python 1 1

6. unverbalized-eval-awareness unverbalized-eval-awareness Public

   Training code for developing a model organism for unverbalized evaluation awareness and for interpretability detection methods experiments

   Python 1