Bohan Tang, DPhil (PhD)

Senior NLP Researcher @ Tencent UK

DPhil in Engineering Science, University of Oxford

I am a Senior NLP Researcher at Tencent UK. I earned my DPhil (PhD) in Engineering Science from the University of Oxford (Machine Learning Research Group), supervised by Prof. Xiaowen Dong, with long-term research co-advising from Prof. Siheng Chen. Prior to Oxford, I completed my Bachelor's degree at Shanghai Jiao Tong University, co-advised by Prof. Siheng Chen and Prof. Ya Zhang.

Google Scholar GitHub LinkedIn

Mentorship & Collaboration: I am always excited to work with self-motivated students (PhD / Master) and researchers on long-horizon and self-improving agents (spanning post-training & RL, agent harnesses, lifelong memory, world models, and high-stakes decision-making). If you are interested in joint projects or academic collaboration aiming for top-tier publications, feel free to reach out via email.

Research Agenda & Vision

"I am an AI researcher working on long-horizon and self-improving agents — building systems that can learn from experience, adapt over time, and tackle increasingly complex tasks."

My research spans both model learning and agent system design, including post-training, reinforcement learning, agent harnesses, memory, tool use, and evaluation. I am particularly interested in how improvements at the model and system levels can work together to enable more capable and reliable agents.

Beyond general agent research, I am interested in applying these systems to real-world forecasting and decision-making, especially in settings where agents must reason over complex, evolving information and support high-stakes decisions.

1. Model Learning & Post-Training

Developing post-training paradigms, RL objectives, and action semantics that prevent compounding errors and support trajectory-level self-improvement.

Post-Training • RL • Action Semantics

2. Agent System Design & Harnesses

Architecting lifelong memory synthesis, generative GUI world models, tool use, agent harnesses, and controllable evaluation frameworks.

Memory • World Models • Harness

3. Forecasting & High-Stakes Decisions

Applying long-horizon agents to settings where systems must reason over complex, evolving information and support high-stakes decision-making.

Forecasting • Evolving Info • High-Stakes

Recent Updates

Oct 2026 MemoryForge accepted to EMNLP 2026! Introduces lifelong memory synthesis for human-like LLM agents.
Jun 2026 Paper on Action Semantics Learning accepted to CVPR 2026 Findings.
May 2026 Honored with the ICML 2026 Gold Reviewer Award.
Apr 2026 Paper ViMo presented at ICLR 2026 as a generative visual GUI world model for app agents.
Feb 2026 DPhil degree in Engineering Science officially conferred at the University of Oxford following minor thesis corrections.
Dec 2025 Joined Tencent UK (London) as a Senior NLP Researcher, leading research on RSI AI agents for long-horizon tasks.
Nov 2025 Passed DPhil Viva examination at the University of Oxford with minor thesis corrections.

Selected Publications

* Equal contribution / Co-first author. Complete list of 19+ papers available on Google Scholar.

Google Scholar

MemoryForge: Synthesize Lifelong Memory for Human-Like LLM Agents

EMNLP 2026

Bohan Tang, Y. Guo

The 2026 Conference on Empirical Methods in Natural Language Processing

Paper

ViMo: A Generative Visual GUI World Model for App Agents

ICLR 2026

D. Luo*, Bohan Tang*, K. Li, G. Papoudakis, J. Song, S. Gong, J. Hao, J. Wang, K. Shao

International Conference on Learning Representations

OpenReview

Beyond Syntax: Action Semantics Learning for App Agents

CVPR 2026

Bohan Tang*, D. Luo*, J. Liu, J. Chen, S. Gong, J. Hao, J. Wang, K. Shao

IEEE/CVF Conference on Computer Vision and Pattern Recognition Findings

Paper

Synthesizing Post-Training Data for LLMs through Multi-Agent Simulation

ACL 2025

S. Tang*, X. Pang*, Z. Liu*, Bohan Tang*, R. Ye, T. Jin, X. Dong, Y. Wang, S. Chen

Annual Meeting of the Association for Computational Linguistics (Main Track)

ACL Anthology

Training-Free Message Passing for Learning on Hypergraphs

ICLR 2025

Bohan Tang, Z. Liu*, K. Jiang*, S. Chen, X. Dong

International Conference on Learning Representations

Paper

Collaborative Uncertainty Benefits Multi-Agent Multi-Modal Trajectory Forecasting

IEEE TPAMI

Bohan Tang*, Y. Zhong*, C. Xu, W.-T. Wu, U. Neumann, Y. Wang, Y. Zhang, S. Chen

IEEE Transactions on Pattern Analysis and Machine Intelligence (2023) / NeurIPS 2021

IEEE TPAMI

Experience & Education

Senior NLP Researcher — Tencent UK

Dec 2025 – Present | London, UK

Leading research on self-evolving AI agents for long-horizon tasks. Mentoring research interns on RSI agent systems, post-training, and AI4AI.

DPhil in Engineering Science — University of Oxford

Oct 2021 – Feb 2026 | Oxford, UK

Machine Learning Research Group. Supervised by Prof. Xiaowen Dong, with research co-advising from Prof. Siheng Chen. Funded by Departmental Research Studentship, Department of Engineering Science, University of Oxford.

BSc in Information Engineering — Shanghai Jiao Tong University

Sep 2017 – Jun 2021 | Shanghai, China

Graduated as University Outstanding Graduate. Co-advised by Prof. Siheng Chen and Prof. Ya Zhang.

Honors & Academic Service

Honors & Awards

ICML Gold Reviewer Award (2026)
NeurIPS Top Reviewer Award (2024)
Departmental Research Studentship, University of Oxford (2021–2025)
SJTU Outstanding Graduate (2021)

Academic Service

Area Chair: ICML GFM Workshop (2026), NeurIPS SEA Workshop (2025).
Journal Reviewer: IEEE-TPAMI, IEEE-TNNLS, IEEE-TSIPN, TMLR.
Conference Reviewer: NeurIPS, ICLR, ICML, CVPR, AAAI, AISTATS, ICASSP.