arxiv:2606.05922
Wenbo Pan
wenbopan
AI & ML interests
Make interesting and accessible models & datasets.
Recent Activity
authored a paper 3 days ago
Retrospective Harness Optimization: Improving LLM Agents via Self-Preference over Trajectory Rollouts authored a paper 6 days ago
The Hidden Dimensions of LLM Alignment: A Multi-Dimensional Analysis of Orthogonal Safety Directions updated a dataset 6 days ago
wenbopan/safety-residual-space