Token-weighted Direct Preference Optimization with Attention
Quick Take
Token-weighted DPO enhances preference optimization by using attention for token importance assessment.
Key Points
- Introduces Token-weighted DPO (TwDPO) for robust training.
- AttentionPO uses LLM's attention to estimate token weights.
- Significantly outperforms existing methods on multiple benchmarks.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →arXiv cs.CL·Leyao Wang, Yanan He, Peng Chen, Asaf Yehudai, Yixin Liu, Rex Ying, Michal Shmueli-Scheuer, Arman Cohan
2d ago FeaturedOriginal
Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents?
AI Summary
The reliability of LLM judges for evaluating deep research agents is critically assessed using the REFLECT benchmark.