8 papers
Memorizing Who, Knowing How: Benchmarking Personal and Procedural Memory for Persistent Assistants
Maosheng Qin, Qi Zhang, Mingxuan Xia +5
Memory systems for persistent personal agents must move beyond isolated memories, maintaining both personal memory about whom they serve and procedural memory about how to act. Cru…
PreSO: Privileged Representation Isolation for On-Policy Self-Distillation
Yuhang Yang, Mingxuan Xia, Chao Ye +6
On-policy self-distillation (OPSD) uses a self-teacher conditioned on privileged information to turn the sparse outcome signal of reinforcement learning with verifiable rewards (RL…
CriPO: Enhancing Rubric-based RL via Self-Distillation
Mingxuan Xia, Yuhang Yang, Chao Ye +8
Rubric-based Reinforcement Learning (RL) has recently shown promise in improving Large Language Models (LLMs) on open-ended tasks. A widely recognized limitation of rubric-based RL…
OPRD: On-Policy Representation Distillation
Shenzhi Yang, Guangcheng Zhu, Bowen Song +8
On-policy distillation (OPD) supervises the student exclusively in the output space by matching next-token distributions. This paradigm suffers from two limitations: (i) a high-var…
CrowdAgent: Multi-Agent Managed Multi-Source Annotation System
Maosheng Qin, Renyu Zhu, Mingxuan Xia +8
High-quality annotated data is a cornerstone of modern Natural Language Processing (NLP). While recent methods begin to leverage diverse annotation sources-including Large Language…
Prompt Candidates, then Distill: A Teacher-Student Framework for LLM-driven Data Annotation
Mingxuan Xia, Haobo Wang, Yixuan Li +4
Recently, Large Language Models (LLMs) have demonstrated significant potential for data annotation, markedly reducing the labor costs associated with downstream applications. Howev…
Towards Robust Incremental Learning under Ambiguous Supervision
Rui Wang, Mingxuan Xia, Chang Yao +4
Traditional Incremental Learning (IL) targets to handle sequential fully-supervised learning problems where novel classes emerge from time to time. However, due to inherent annotat…
Revisiting the Robust Generalization of Adversarial Prompt Tuning
Fan Yang, Mingxuan Xia, Sangzhou Xia +2
Understanding the vulnerability of large-scale pre-trained vision-language models like CLIP against adversarial attacks is key to ensuring zero-shot generalization capacity on vari…