12 papers · 1 filter
OP-Print: Tracing Membership Imprints in On-Policy Self-Distillation
Qi Li, Xinchao Wang
On-policy distillation has been widely adopted for alleviating the distribution mismatch of off-policy distillation. Self-distillation further removes the need for an external teac…
Jeva: System One Models from First Principles
Qi Li, Xinchao Wang
System one models like Jev enable direct, low-latency decision-making. Yet recent open-source practice of predicting an answer from serialized text does not, by itself, respect the…
PRISM: PRIvileged Sibling-guided Multi-Agent Topology Design
Phuong Tuan Dat, Qi Li, Xinchao Wang
Multi-agent systems (MAS) based on large language models (LLMs) have become a powerful way to address complex problems, and their effectiveness depends critically on the collaborat…
BadWAM: When World-Action Models Dream Right but Act Wrong
Qi Li, Xingyi Yang, Xinchao Wang
World-action models (WAMs) are emerging as a promising foundation for embodied control: rather than predicting actions alone, they learn representations that couple action generati…
dOPSD: On-Policy Self-Distillation for Diffusion Language Models
Phuong Tuan Dat, Qi Li, Xinchao Wang
Diffusion large language models (dLLMs) generate text by iteratively denoising a masked sequence, offering a parallel alternative to autoregressive models, but eliciting strong rea…
ViMU: Benchmarking Video Metaphorical Understanding
Qi Li, Xinchao Wang
Any new medium, once it emerges, is used for more than the transmission of overt content alone. The information it carries typically operates on two levels: one is the content dire…
On-Policy Self-Evolution via Failure Trajectories for Agentic Safety Alignment
Bo Yin, Qi Li, Xinchao Wang
Tool-using LLM agents fail through trajectories rather than only final responses, as they may execute unsafe tool calls, follow injected instructions, comply with harmful requests,…
AutoMIA: Improved Baselines for Membership Inference Attack via Agentic Self-Exploration
Ruhao Liu, Weiqi Huang, Qi Li +1
Membership Inference Attacks (MIAs) serve as a fundamental auditing tool for evaluating training data leakage in machine learning models. However, existing methodologies predominan…
Anatomy of a Lie: A Multi-Stage Diagnostic Framework for Tracing Hallucinations in Vision-Language Models
Lexiang Xiong, Qi Li, Jingwen Ye +1
Vision-Language Models (VLMs) frequently "hallucinate" - generate plausible yet factually incorrect statements - posing a critical barrier to their trustworthy deployment. In this…
Self-Purification Mitigates Backdoors in Multimodal Diffusion Language Models
Guangnian Wan, Qi Li, Gongfan Fang +2
Multimodal Diffusion Language Models (MDLMs) have recently emerged as a competitive alternative to their autoregressive counterparts. Yet their vulnerability to backdoor attacks re…
Sponge Tool Attack: Stealthy Denial-of-Efficiency against Tool-Augmented Agentic Reasoning
Qi Li, Xinchao Wang
Enabling large language models (LLMs) to solve complex reasoning tasks is a key step toward artificial general intelligence. Recent work augments LLMs with external tools to enable…
Refinement Provenance Inference: Detecting LLM-Refined Training Prompts from Model Behavior
Bo Yin, Qi Li, Runpeng Yu +1
Instruction tuning increasingly relies on LLM-based prompt refinement, where prompts in the training corpus are selectively rewritten by an external refiner to improve clarity and…