Song Mei
- Papers
- 4
Cite
Notes
Only stored in your browser.
4papers
Authored papers
4Improving LLM Safety Alignment with Dual-Objective Optimization
arXiv 2025
Active-Dormant Attention Heads: Mechanistically Demystifying Extreme-Token Phenomena in LLMs
arXiv 2024
Simplicity Prevails: Rethinking Negative Preference Optimization for LLM Unlearning
arXiv 2024
Transformers as Decision Makers: Provable In-Context Reinforcement Learning via Supervised Pretraining
arXiv 2023
Affiliations
No known affiliations.
Frequent co-authors
10from 4 papers