7 papers
When Is a Task Vector Enough? An Empirical Theory of Implicit Multimodal ICL
Jiaqian Li
Implicit multimodal in-context learning compresses demonstrations into internal interventions, ranging from static task vectors to query-conditioned transformations and attention r…
The Flip Side of RLHF: On-Policy Feedback for Reward Model Self-Supervised Improvement
Xiaobo Wang, Tong Wu, Min Tang +3
Building strong reward models (RMs) for language model alignment is bottlenecked by the cost and difficulty of acquiring diverse and reliable preference data from human annotation…
TRACES: Proactive Safety Auditing for Multi-Turn LLM Agents via Trajectory-State Modeling
Jiaqian Li, Yanshu Li, Boxuan Zhang +2
LLM agents increasingly operate through multi-turn tool use and environment interaction, where safety risks often emerge from intermediate steps long before they surface in the fin…
Personalize Your Large Vision-language Models With In-context Prompt Tuning
Yanshu Li, Jiaqian Li, Kuai Yu +4
Large vision-language models (LVLMs) have demonstrated strong general multimodal capability and are increasingly deployed in downstream systems. This trend has driven growing inter…
Steering Vector Fields for Context-Aware Inference-Time Control in Large Language Models
Jiaqian Li, Yanshu Li, Kuan-Hao Huang
Steering vectors (SVs) offer a lightweight way to control large language models (LLMs) at inference time by shifting hidden activations, providing a practical middle ground between…
Train Once, Reuse Everywhere: Generalizable Implicit In-Context Learning by Routing Attention
Jiaqian Li, Yanshu Li, Ligong Han +2
Implicit in-context learning (ICL) has newly emerged as a promising paradigm that simulates ICL behaviors in the representation space of large language models (LLMs), aiming to att…