4 papers
Trajectory-Relative Hindsight Distillation for Agentic Reinforcement Learning
Haoyu Zheng, Yun Zhu, Qing Wang +1
Recent agentic reinforcement learning methods use hindsight to complement sparse outcome rewards. However, a completed rollout can yield many such signals, leaving their appropriat…
MAIGO: Mitigating Lost-in-Conversation with History-Cleaned On-Policy Self-Distillation
Haoyu Zheng, Yun Zhu, Shu Yuan +5
Large language models often solve tasks from a fully specified prompt but degrade when the same requirements unfold over multiple turns, known as the lost-in-conversation (LiC) gap…
PILOT: Planning via Internalized Latent Optimization Trajectories for Large Language Models
Haoyu Zheng, Yun Zhu, Yuqian Yuan +4
Strategic planning is critical for multi-step reasoning, yet compact Large Language Models (LLMs) often lack the capacity to formulate global strategies, leading to error propagati…
Scientific Graphics Program Synthesis via Dual Self-Consistency Reinforcement Learning
Juekai Lin, Yun Zhu, Honglin Lin +6
Graphics Program Synthesis is pivotal for interpreting and editing visual data, effectively facilitating the reverse-engineering of static visuals into editable TikZ code. While Ti…