6 papers
Evo: Evolving the Evolver through Weight-Harness Co-Evolution
Chejian Xu, Shujian Zhang, Wenxuan Zhou +5
Self-evolving agents are increasingly used to improve language model systems without human intervention through either weight updates or harness improvements. Existing approaches t…
Steering LLMs for Culturally Localized Generation
Simran Khanuja, Hongbin Liu, Shujian Zhang +4
LLMs are deployed globally, yet produce responses biased towards cultures with abundant training data. Existing cultural localization approaches such as prompting or post-training…
MUSIC: MUlti-Step Instruction Contrast for Multi-Turn Reward Models
Wenzhe Li, Shujian Zhang, Wenxuan Zhou +5
Evaluating the quality of multi-turn conversations is crucial for developing capable Large Language Models (LLMs), yet remains a significant challenge, often requiring costly human…
Fantastic Reasoning Behaviors and Where to Find Them: Unsupervised Discovery of the Reasoning Process
Zhenyu Zhang, Shujian Zhang, John Lambert +6
Despite the growing reasoning capabilities of recent large language models (LLMs), their internal mechanisms during the reasoning process remain underexplored. Prior approaches oft…
Eliciting Behaviors in Multi-Turn Conversations
Jing Huang, Shujian Zhang, Lun Wang +3
Identifying specific and often complex behaviors from large language models (LLMs) in conversational settings is crucial for their evaluation. Recent work proposes novel techniques…
DPO Unchained: Your Training Algorithm is Secretly Disentangled in Human Choice Theory (and its Loss' Convexity is Dispensable)
Wenxuan Zhou, Shujian Zhang, Brice Magdalou +4
Normative theories allow one to elicit key parts of a ML algorithm from first principles, which is crucial at a time of championed scrutiny for ML work. Direct Preference Optimizat…