EMNLP 2026 Unreleased manuscript
MessageSieve
Runtime Message Pruning for Accurate and Efficient LLM-Based Multi-Agent Communication
ZHEJIANG UNIVERSITY
PhD student · Research on reasoning and autonomous agents
I am a PhD student at Zhejiang University, since 2023, advised by Prof. Hehe Fan. I study how language models can organize their reasoning, investigate difficult questions with evidence, and collaborate as agents.
I collaborate with Ant Group on research (September 2025 – October 2026) and serve as a reviewer for ACL 2026 and ICLR 2027.
First-authored and coauthored work.
EMNLP 2026 Unreleased manuscript
Runtime Message Pruning for Accurate and Efficient LLM-Based Multi-Agent Communication
SIGGRAPH Asia 2026 Accepted
Tree-Graph Guided Multi-Agent Text-to-SVG Generation

arXiv 2026 First author
Answering Highly Complex Questions with Large Language Models through Super Deep and Super Wide Research
A benchmark for complex autonomous research through deep investigation and broad evidence gathering.

ICLR 2026 First author
Structured Reasoning for LLMs: A Unified Framework for Efficiency and Explainability
Explicit reasoning-step structure for efficient and explainable language models.
Dataset and project materials available; training code and checkpoints not yet released.

CVPR 2026 Second author
Look at What You'll Lose before Pruning with Hierarchical Token Selection
Efficient vision-language inference through hierarchical token selection and information merging.
Project information available; implementation not yet released.

ACL 2024 First author
A Graph-Based Multi-Agent Framework for Coordinating Complex Task Dependencies in Minecraft
Coordinating collaborative agents through explicit task graphs in Minecraft.
Research collaboration
ACL 2026 · ICLR 2027