Pinned Loading

1. Phi Phi Public

   [EMNLP 2025 Oral] Phi: Preference Hijacking in Multimodal Large Language Models at Inference Time

   Python 11

2. zero-cot-probe zero-cot-probe Public

   Zero-CoT Probe (ZCP) for detecting evasive data contamination in LLMs. Paper Link: http://arxiv.org/abs/2605.21856

   Python 1

3. awesome-jev-robustness awesome-jev-robustness Public

   Tests, calibration audits and failure-mode studies of Jev (TypeSafe System One): jaggedness, consistency, injection, abstention.

   5 2

4. Skill-Transfer Skill-Transfer Public

   Code for Skill Transfer

   Python 1