10 papers · 1 filter
Beyond the Teacher: Understanding and Preserving Distillation Overshoot in Agentic On-Policy Distillation
Jike Zhong, Ming Li, Yuxiang Lai +9
On-policy distillation (OPD) trains agents on their own trajectories using supervision from a stronger teacher. We uncover a counterintuitive phenomenon in multi-turn agentic tasks…
DeLoop: Repeated Actions as Self-Supervision for Long-Horizon Agents
Tianle Chen, Pushkar Singh, Ketan Todi +8
Language-model agents that run long tasks often redo work they have already done: they re-issue a tool call they made earlier and get back exactly the same observation, a step that…
DTDD: Divergence-Triggered Dynamic Distillation for Reliable On-Policy Supervision
Qijia He, Zequn Li, Spencer Luo +6
On-policy distillation (OPD) provides dense token-level supervision on student rollouts, but this supervision can become unreliable when the student and teacher diverge, either bec…
Improving LLM Collaboration via Multi-Agent Preference Learning
Shuo Liu, Xinzichen Li, Tianle Chen +1
Several works have explored multi-agent reinforcement learning (MARL) in LLM collaboration. However, constructing reliable rewards is difficult in practice, as complete and accurat…
Fewer Tokens, Smaller Cache: Reward-Coordinated Efficient Reasoning
Qiyuan Zhu, Dezhi Li, Pengyu Cheng +8
Large Reasoning Models (LRMs) excel on complex tasks through long chain-of-thought (CoT) reasoning, but their lengthy intermediate steps cause severe overthinking that inflates inf…
The Curse of Helpfulness: Inverse Scaling Law in Robustness to Distractor Instructions via DistractionIF
Zeli Su, Zhankai Xu, Tianlei Chen +4
Large Language Models (LLMs) are increasingly deployed in agentic and retrieval-augmented generation (RAG) systems, where they must execute user-specified tasks over externally pro…
Counteraction-Aware Multi-Teacher On-Policy Distillation for General Capability Recovery with Domain Preservation
Tianlei Chen, Jiao Ou, Ziyuan Liu +3
Domain specialization can improve LLM behavior, but often weakens the general capabilities inherited from the original model. Recent Multi-Teacher On-Policy Distillation (MOPD) pip…
A Systematic Study of Cross-Modal Typographic Attacks on Audio-Visual Reasoning
Tianle Chen, Deepti Ghadiyaram
As audio-visual multi-modal large language models (MLLMs) are increasingly deployed in safety-critical applications, understanding their cross-modal vulnerabilities is essential. T…
VLM-UQBench: A Benchmark for Modality-Specific and Cross-Modality Uncertainties in Vision Language Models
Chenyu Wang, Tianle Chen, H. M. Sabbir Ahmad +2
Uncertainty quantification (UQ) is vital for ensuring that vision-language models (VLMs) behave safely and reliably. A central challenge is to localize uncertainty to its source, d…
Learning Decentralized LLM Collaboration with Multi-Agent Actor Critic
Shuo Liu, Tianle Chen, Ryan Amiri +1
Recent work has explored optimizing LLM collaboration through Multi-Agent Reinforcement Learning (MARL). However, most MARL fine-tuning approaches rely on predefined execution prot…