Summary

I am Dayuan Fu, a graduate student at PRIS-NLP Group at Beijing University of Posts and Telecommunications (BUPT), supervised by Prof. Weiran Xu. I'm visiting GAIR now, and I visited TsinghuaC3I group from 2022.10 to 2024.8. My research interests primarily focus on LLM Agent's reasoning, planning, and decision-making ability and memory strategy, which can make LLM more universal (i.e. AGI). I have published several papers at prominent NLP conferences, including ICLR, EMNLP, CIKM, and NAACL.

Education

• M.S. in Dept. of Artificial Intelligence

  2023.9-2026.6
  Beijing University of Posts and Telecommunications (BUPT)
• B.S. in Dept. of Electronic Engineering

  2019.9-2023.6
  Beijing University of Posts and Telecommunications (BUPT)

  GPA: 93.18/100 (3.88/4.0)

Visiting Experience

• C3I Group

  2022.10 - 2024.8
  Tsinghua University
• GAIR

  2024.9 - Now
  SII & SJTU

Industry Experience

• Meituan, NLP Center

  2024.1 - 2025.3
  Algorithm Intern

  Research Intern on Agent and data synthesis.

Publications

• DeepResearcher: Scaling Deep Research via Reinforcement Learning in Real-world Environments

  2025
  View Publication
• AgentRefine: Enhancing Agent Generalization through Refinement Tuning

  2025
  ICLR 2025

  View Publication
• MSI-Agent: Incorporating Multi-Scale Insight into Embodied Agents for Superior Planning and Decision-Making

  2024
  EMNLP 2024 Main

  View Publication
• How Do Your Code LLMs Perform? Empowering Code Instruction Tuning with High-Quality Data

  2024
  EMNLP 2024 Main

  View Publication
• CS-Bench: A Comprehensive Benchmark for Large Language Models towards Computer Science Mastery

  2024
  ICLR 2025

  View Publication
• PreAct: Prediction Enhances Agent's Planning Ability

  2024
  Coling 2025

  View Publication
• On Large Language Models' Hallucination with Regard to Known Facts

  2024
  NAACL 2024 Main

  View Publication
• DivTOD: Unleashing the Power of LLMs for Diversifying Task-Oriented Dialogue Representations

  2024
  NAACL 2024 Findings

  View Publication
• A multi-task semantic decomposition framework with task-specific pre-training for few-shot ner

  2023
  CIKM 2023

  View Publication
• A prototypical semantic decoupling method via joint contrastive learning for few-shot named entity recognition

  2023
  ICASSP 2023

  View Publication
• Revisit out-of-vocabulary problem for slot filling: A unified contrastive framework with multi-level data augmentationss

  2023
  ICASSP 2023

  View Publication
• Semi-supervised knowledge-grounded pre-training for task-oriented dialog systems

  2022
  SereTOD2022, EMNLP 2022 Workshop

  View Publication

Skills

Languages

Honors & Awards

• Excellent First-class Scholarship for Master Students, BUPT

  2023, 2024
• the 1st Award on SereTOD Challenge 2022 track 2

  2022
• National Scholarship(Top 1%)

  2021
• First Prize in The Chinese Mathematics Competitions

  2020, 2021, 2022