Qineng Wang

CS Ph.D. Student, Northwestern University

my_pic.jpg

Email: qinengwang2029 # u dot northwestern dot edu

Bibliography

Hi there! I’m Qineng Wang (王启能). I am a first-year CS Ph.D. student at Northwestern University. I am very fortunate to work with Prof. Manling Li🌟. I got my B.Eng. in Computer Science and Technology from Zhejiang University.

Currently, I am a research intern at SVL Lab @ Standford University, working with Prof. Li Fei-Fei and Prof. Jiajun Wu on spatial intelligence and embodied agents.

Research Interests

My long-term research goal is to empower foundation models with strong spatial intelligence and reasoning capabilities to drive embodied agents toward achieving spatial-aware, long-horizon goals. I am broadly interested in multi-modal, interactive, and reasoning-capable agents, with a focus on the following research directions:

Embodied Agentic Planning and Dynamics Understanding

Leveraging foundation models to understand environment dynamics and support embodied agents in planning, decision-making, and long-horizon goal completion (EAI, EmbodiedBench).

Spatial Intelligence and Mental Modeling

Investigating and developing methods to equip foundation models with spatial mental modeling capabilities, enabling them to reason about object consistency,object relations, physical affordances, and spatial configurations (MindCube).

Reasoning Agents with Foundation Models

Exploring how (1) multi-agent discussions can be integrated into foundation models to support collaborative reasoning (CMD), and (2) reinforcement learning can be applied to foundation models in multi-turn agentic settings (RAGEN, VAGEN).

Reasoning in Continuous Representational Spaces

Investigating reasoning mechanisms over continuous latent spaces (e.g., embeddings or trajectories), to enable foundation models to make abstract, spatial, and semantic inferences beyond symbolic logic.

News

May 2025 EmbodiedBench got accepted by ICML 2025.
Mar 2025 We released RAGEN, the first multi-turn reinforcement learning framework for LLM agents.
Mar 2025 Will co-organize a workshop on Foundation Models Meet Embodied Agents at CVPR 2025.
Nov 2024 EAI won the Best Paper Award at SoCal NLP 2024.
Sep 2024 EAI got accepted by NeurIPS 2024 as an oral presentation.
Jun 2024 Graduated as an outstanding undergrad from Zhejiang University.
May 2024 CMD got accepted by ACL 2024.
Mar 2024 Will join Northwestern University as a CS PhD student, working with Prof. Manling Li.

Selected Publications

2025

  1. arXiv
    ragen.png
    Ragen: Understanding self-evolution in llm agents via multi-turn reinforcement learning
    Zihan Wang†, Kangrui Wang†, Qineng Wang†, Pingyue Zhang†, Linjie Li†, Zhengyuan Yang, Xing Jin, Kefan Yu, Minh Nhat Nguyen, Licheng Liu, Eli Gottlieb, Yiping Lu, Kyunghyun Cho, Jiajun Wu, Li Fei-Fei, Lijuan Wang, Yejin Choi, and Manling Li
    May 2025
    arXiv Code Website

2024

  1. NeurIPS
    eai.png
    Embodied Agent Interface: Benchmarking LLMs for Embodied Decision Making
    Manling Li†, Shiyu Zhao†, Qineng Wang†, Kangrui Wang†, Yu Zhou†, Sanjana Srivastava, Jem Gokmen, Tony Lee, Li Erran Li, Ruohan Zhang, Weiyu Liu, Percy Liang, Li Fei-Fei, Jiayuan Mao, and Jiajun Wu
    In Proceedings of the 38th International Conference on Neural Information Processing Systems (NeurIPS, D&B Track), Dec 2024
    NeurIPS Oral Presentation | SoCal NLP Best Paper Award
    arXiv Code Website
  2. ACL
    cmd.png
    Rethinking the Bounds of LLM Reasoning: Are Multi-Agent Discussions the Key?
    Qineng Wang†, Zihao Wang†, Ying Su, Hanghang Tong, and Yangqiu Song
    In Proceedings of the 62th Annual Meeting of the Association for Computational Linguistics (ACL), Aug 2024
    arXiv

Services

Workshop Organization: FMEA@CVPR 2025 (Program Chair)
Conference Reviews: NeurIPS 2025; FMEA@CVPR 2025; KnowLM@ACL 2024
External Journal Reviews: TACL 2024