About

Hi! I am a Ph.D. student in the School of Computer Science at Harbin Institute of Technology, supervised by Wangmeng Zuo and Lei Bai. Previously, I was a research intern at the Shanghai Artificial Intelligence Laboratory, where I had the pleasure of working closely with Yiran Qin and Zhiheng Xi.

My research interests center on embodied manipulation, LLM agents, and multimodal reasoning, with a recent focus on reinforcement learning and foundation models for embodied agents.

  • Publications at NeurIPS 2025, ICLR 2025, ICML 2026, ACL 2026 (Oral), and TMLR.
  • Research spans LLM agents, reinforcement learning, multimodal reasoning, and embodied foundation models.
  • Building large-scale datasets and benchmarks for multimodal and scientific reasoning.

Research Focus

Embodied Manipulation

Foundation models and world models that ground perception, language, and action for embodied agents.

LLM Agents

Training and evaluating agentic systems that search, plan, collaborate, and adapt across dynamic environments.

Multimodal Reasoning

Improving how vision-language models understand ambiguous instructions, typography, and visual context.

Education

  1. Harbin Institute of Technology & AI Lab
    Ph.D. Student, advised by Prof. Wangmeng Zuo and Prof. Lei Bai
    Sep. 2025 - present
  2. Harbin Institute of Technology
    Bachelor in Software Engineering
    Sep. 2021 - Jul. 2025

Experience

  1. Xiaomi Robotics
    Research Intern ยท Top Talent Program
    Oct. 2026 - Present
  2. Xinzhi
    Research Intern
    Feb. 2026 - Sep. 2026
  3. Shanghai AI Laboratory
    Research Intern
    Oct. 2024 - Feb. 2026

News

  • We released $N_0$-Foundation, $N_0$-VTLA, and $N_0$-TWAM, technical reports towards tactile intelligence for robotic manipulation!
  • The Landscape of Agentic Reinforcement Learning for LLMs: A Survey is accepted by TMLR!
  • Our study on scaling behaviors of reinforcement learning for LLMs is accepted by ACL 2026 as an Oral!
  • SciAgentGym is accepted by ICML 2026!
  • Does Reinforcement Fine-Tuning Improve Generalization of LLM Agents? An Empirical Study is accepted by ICML 2026!
  • BMMR is accepted by NeurIPS 2025!
  • I started my Ph.D. at Harbin Institute of Technology, jointly working with Shanghai AI Laboratory.
  • Visual-O1 is accepted by ICLR 2025!

Selected Publications

All publications
* equal contribution · # corresponding author
Tech Report

$N_0$-Foundation: Towards the Age of Tactile Intelligence

NeoteAI Team, Fudan TEAI Team
arXiv coming soon. 2026 Tech Report Core Contributor

$N_0$-Foundation is a tactile-centric foundation for embodied manipulation that integrates tactile sensing hardware, large-scale multimodal data, hardware-agnostic tactile representation learning, and standardized evaluation into one system. Durable vision-based tactile sensors and $N_0$-TacUMI provide a synchronized acquisition stack; NeoData supplies more than 30,000 hours and 1.4M episodes of synchronized visual-tactile interaction across six embodiments and 450+ tasks (with OpenNeoData, a 5,000-hour open-source subset); NeoForce learns transferable contact features from device-independent three-axis force fields; and the NeoReal and NeoSim suites standardize evaluation for contact-rich manipulation in the physical world and simulation.

Report Dataset Code Website
Tech Report

$N_0$-VTLA: Scaling Vision-Tactile-Language-Action Model with Latent Tactile Tokens

NeoteAI Team, Fudan TEAI Team
2026 Tech Report Core Contributor

$N_0$-VTLA is a vision-tactile-language-action (VTLA) foundation model for fine-grained contact-rich manipulation. To our knowledge, it is the first VTLA model pretrained on tactile data at scale, learning broad contact priors from NeoData, our large-scale visuo-tactile robot dataset. A predictive tactile pathway distills the contact patterns learned at scale into fine motion adjustments, and ALTER, an advantage-conditioned offline RL method, further improves policies from stored deployment data. $N_0$-VTLA wins all nine real-robot NeoReal tasks and reaches 63.8% mean success on a twenty-task simulation suite, against 44.0% for the strongest baseline.

Report Code Website
Tech Report

$N_0$-TWAM: Scaling Tactile-Native World-Action Model for Contact-Rich Manipulation

NeoteAI Team, Fudan TEAI Team
2026 Tech Report Core Contributor

$N_0$-TWAM is a tactile-native world-action model for contact-rich manipulation that jointly predicts future vision and future contact. To our knowledge, it is the first tactile world-action model trained at large scale, pre-trained with visuo-tactile joint training over tactile-rich demonstrations spanning six embodiments and 450 tasks. An asymmetric Mixture-of-Transformers architecture pairs a full-width video expert with slim action and tactile experts, so touch is both predicted and observed while inference stays real-time. Evaluations on real and simulated benchmarks demonstrate strong capability across contact-rich tasks and clear benefits from data scaling for precise tactile and action prediction.

Report Code Website
PreprintCoEnv: Driving Embodied Multi-Agent Collaboration via Compositional Environment

CoEnv: Driving Embodied Multi-Agent Collaboration via Compositional Environment

Li Kang*, Yutao Fan*, Rui Li, Heng Zhou, Yiran Qin, Zhemeng Zhang, Songtao Huang, Xiufeng Song, Zaibin Zhang, Bruno N.Y. Chen, Zhenfei Yin, Dongzhan Zhou, Wangmeng Zuo, Lei Bai
2026 Preprint

CoEnv addresses coordination challenges in multi-agent robotic systems by introducing a compositional environment that integrates real-world and simulation components. The framework operates through three stages: digitizing physical workspaces, employing language models for action planning, and transferring validated strategies to real robots with collision safeguards. We demonstrate the effectiveness of our approach on complex multi-arm manipulation tasks, proposing a new paradigm for multi-agent embodied AI.

Paper
ICML 2026Does Reinforcement Fine-Tuning Improve Generalization of LLM Agents? An Empirical Study

Does Reinforcement Fine-Tuning Improve Generalization of LLM Agents? An Empirical Study

Zhiheng Xi*, Xin Guo*, Jiaqi Liu*, Jiazheng Zhang*, Yutao Fan, Zhihao Zhang, Shichun Liu, Mingxu Chai, Xiaowei Shi, Yitao Zhai, Xunliang Cai, Tao Gui, Qi Zhang, Xuanjing Huang
International Conference on Machine Learning (ICML) 2026 Regular

We investigate whether reinforcement fine-tuning effectively enables LLM agents to generalize across scenarios. Our study examines three dimensions: within-environment generalization across task difficulty, cross-environment transfer to unseen environments, and sequential multi-environment training to assess both transfer and forgetting. Findings reveal that while RFT performs well within familiar environments, it shows weaker transfer to unseen environments, correlating with shifts in semantic priors and observation/action interfaces. Sequential and mixture training approaches demonstrate promise for improving overall generalization while maintaining prior knowledge.

Paper
PreprintCoEditor++: Instruction-based Visual Editing via Cognitive Reasoning

CoEditor++: Instruction-based Visual Editing via Cognitive Reasoning

Minheng Ni, Yutao Fan, Zhengyuan Yang, Yeli Shen, Yuxiang Wei, Yaowen Zhang, Lijuan Wang, Lei Zhang, Wangmeng Zuo
2026 Preprint

CoEditor++ is a training-free framework for instruction-based image editing that decomposes editing into "what to edit" and "how to edit" through two cognitive stages with a reflective self-selection mechanism. Built from publicly available components, CoEditor++ achieves competitive results on SmartEdit and AltBear compared to both open-source and proprietary models including GPT-4o, with particularly strong performance in visual consistency while maintaining instruction-following capabilities. We show the effectiveness of CoEditor++ benefits from its structured cognitive design rather than any specific model component.

Paper
Under ReviewBetter know nothing than half-know anything: A Precise and Efficient Dataset for Scientific Reasoning in Language Models

Better know nothing than half-know anything: A Precise and Efficient Dataset for Scientific Reasoning in Language Models

Yutao Fan*, Yizhou Wang*, Zhiheng Xi*, Lintao Wang, Jianyu Wu, Aoran Wang, Guanyu Li, Jiaqi Liu, Pengze Li, Heng Zhou, Jiayang Li, Wangmeng Zuo, Lei Bai, Shixiang Tang, Philip Torr, Zhenfei Yin
Under review. 2025

Large Language Models (LLMs) have achieved remarkable progress in reasoning tasks, i.e., coding and mathematics. However, their ability to perform scientific reasoning remains significantly limited, probably hampered by the scarcity of high-quality scientific reasoning datasets. Existing approaches either rely on LLM-generated synthetic data (suffering from noise and hallucinations) or human-compiled documents (facing scarcity and non-standardization). In this paper, we empirically verify that integrating precise knowledge from original scientific documents with formalized questions and consistent answers can mitigate the need for large-scale data. Based on this insight, we design PreciSci, a pipeline for constructing multi-disciplinary scientific reasoning datasets. This pipeline involves extracting knowledge from reliable sources, refining questions for completeness and precision, applying multi-stage filtering to eliminate redundancy and noise, and refining answers to ensure reliable supervision. Leveraging PreciSci, we build Open-Sci, a precise and knowledge-dense scientific reasoning dataset. Experimental evaluations show that despite Open-Sci being less than one-sixth the size of state-of-the-art scientific reasoning datasets, it enables LLMs to achieve approximately 4.49% better performance across diverse discipline-specific benchmarks.

NeurIPS 2025BMMR: A Large-Scale Bilingual Multimodal Multi-Discipline Reasoning Dataset

BMMR: A Large-Scale Bilingual Multimodal Multi-Discipline Reasoning Dataset

Zhiheng Xi*, Guanyu Li*, Yutao Fan*, Honglin Guo*, Yufang Liu, Xiaoran Fan, Jiaqi Liu, Jingchao Ding, Wangmeng Zuo, Zhenfei Yin, Lei Bai, Tao Ji, Tao Gui, Qi Zhang, Philip Torr, Xuanjing Huang
Annual Conference on Neural Information Processing Systems (NeurIPS) 2025 Poster

In this paper, we introduce BMMR, a large-scale bilingual, multimodal, multi-disciplinary reasoning dataset for the community to develop and evaluate large multimodal models (LMMs). BMMR comprises 110k college-level questions spanning 300 UNESCO-defined subjects, spanning diverse formats-multiple-choice, fill-in-the-blank, and open-ended QA-and sourced from both print and digital media such as books, exams, and quizzes. All data are curated and filtered via a human-in-the-loop and scalable framework, and each instance is paired with a high-quality reasoning path. The dataset is organized into two parts: BMMR-Eval that comprises 20,458 high-quality instances to comprehensively assess LMMs'knowledge and reasoning across multiple disciplines in both Chinese and English; and BMMR-Train that contains 88,991 instances to support further research and development, extending the current focus on mathematical reasoning to diverse disciplines and domains. In addition, we propose the process-based multi-discipline verifier (i.e., BMMR-Verifier) for accurate and fine-grained evaluation of reasoning paths. Extensive experiments on 24 models reveal that (i) even SOTA models (e.g., o3 and Gemini-2.5-Pro) leave substantial headroom on BMMR-Eval; (ii) reasoning models exhibit discipline bias and outperform LMMs only on specific subjects; (iii) open-source models still trail their proprietary counterparts; and (iv) fine-tuning on BMMR-Train narrows this gap. Additionally, we conduct reasoning-chain analyses using BMMR-Verifier and other in-depth studies, uncovering the challenges LMMs currently face in multidisciplinary reasoning.

Paper Homepage Code Dataset
ICLR 2025Visual-O1: Understanding Ambiguous Instructions via Multi-modal Multi-turn Chain-of-thoughts Reasoning

Visual-O1: Understanding Ambiguous Instructions via Multi-modal Multi-turn Chain-of-thoughts Reasoning

Minheng Ni, Yutao Fan, Lei Zhang, Wangmeng Zuo
The International Conference on Learning Representations (ICLR) 2025 Poster

As large-scale models evolve, language instructions are increasingly utilized in multi-modal tasks. Due to human language habits, these instructions often contain ambiguities in real-world scenarios, necessitating the integration of visual context or common sense for accurate interpretation. However, even highly intelligent large models exhibit significant performance limitations on ambiguous instructions, where weak reasoning abilities of disambiguation can lead to catastrophic errors. To address this issue, this paper proposes Visual-O1, a multi-modal multi-turn chain-of-thought reasoning framework. It simulates human multi-modal multi-turn reasoning, providing instantial experience for highly intelligent models or empirical experience for generally intelligent models to understand ambiguous instructions. Unlike traditional methods that require models to possess high intelligence to understand long texts or perform lengthy complex reasoning, our framework does not significantly increase computational overhead and is more general and effective, even for generally intelligent models. Experiments show that our method not only significantly enhances the performance of models of different intelligence levels on ambiguous instructions but also improves their performance on general datasets. Our work highlights the potential of artificial intelligence to work like humans in real-world scenarios with uncertainty and ambiguity.

Paper Code

Honors & Awards

  • Outstanding Student of Harbin Institute of Technology 2025
  • Huawei Scholarship 2024