Zilong Wang

PhD Student at UC San Diego, CSE.

profile.jpg

Welcome! I am a fifth-year PhD student at UC San Diego advised by Prof. Jingbo Shang. I spent wonderful time doing research at Amazon Foundation Modl, Google Cloud AI, Google DeepMind, Google Research, Adobe Research, and Microsoft Research Asia. I received my B.S. in Computer Science from Peking University in 2020, where I was advised by Prof. Xiaojun Wan.

My research spans several areas of natural language processing, including reasoning, information extraction, multimodal learning, and language modeling. I primarily focus on LLM post-training, targeting at aligning LLMs to complex reasoning and planning with knowledge-intensive queries, autonomous agents, and mathematical problem-solving. Currently, I am exploring the collaboration of weak and strong LLMs and the process supervision with Monte Carlo Tree Search.

My earlier research forcused on visually-rich document understanding, where I enabled language models to encode multimodal features and understand the rich contents in documents of various formats, such as forms, receipts, web pages, etc.

news

Sep 1, 2024 Speculative RAG: Enhancing Retrieval Augmented Generation through Drafting New paper alert! Achieve state-of-the-art performance both in accuracy and efficiency for RAG.
Jul 26, 2024 OFFICEBENCH: Benchmarking Language Agents across Multiple Applications for Office Automation New paper alert! Check our latest LLM agent benchmark on the office automation scenario!
May 27, 2024 Debug like a Human: A Large Language Model Debugger via Verifying Runtime Execution Step-by-step has been accepted by ACL 2024 Findings!
Jan 16, 2024 Chain-of-Table: Evolving Tables in the Reasoning Chain for Table Understanding has been accepted by ICLR 2024!
Dec 9, 2023 A Study on Robustness and Reliability of Large Language Model Code Generation got received by AAAI 2024!

selected publications

  1. Speculative RAG: Enhancing Retrieval Augmented Generation through Drafting
    Zilong Wang, Zifeng Wang, Long Le, and 9 more authors
    arXiv preprint arXiv:2407.08223, 2024
  2. OfficeBench: Benchmarking Language Agents across Multiple Applications for Office Automation
    Zilong Wang, Yuedong Cui, Li Zhong, and 4 more authors
    arXiv preprint arXiv:2407.19056, 2024
  3. ACL ’24
    Debug like a Human: A Large Language Model Debugger via Verifying Runtime Execution Step-by-step
    Li Zhong, Zilong Wang**, and Jingbo Shang
    In ACL 2024 (Findings), 2024
  4. ICLR ’24
    Chain-of-Table: Evolving Tables in the Reasoning Chain for Table Understanding
    Zilong Wang, Hao Zhang, Chun-Liang Li, and 8 more authors
    In ICLR 2024, 2024
  5. ACL ’24
    Answer is All You Need: Instruction-following Text Embedding via Answering the Question
    Letian Peng, Yuwei Zhang, Zilong Wang, and 4 more authors
    In ACL 2024, 2024
  6. AAAI ’24
    A Study on Robustness and Reliability of Large Language Model Code Generation
    Li Zhong, and Zilong Wang
    In AAAI 2024, 2024