CZ. Get in touch
Researcher in AI and decision-making Singapore

Chubin Zhang.

Learning through interaction.

I work on learning and decision-making, with recent projects in online reinforcement learning and generative policies.

Explore my research GitHub

01 / About

About me

Chubin Zhang outdoors in Singapore

I am Chubin Zhang (张楚彬), currently a Research Assistant in Prof. Bo An's group at Nanyang Technological University.

My recent work has explored online reinforcement learning and expressive generative policies. More broadly, I am interested in how learning systems make decisions and adapt through interaction.

For research conversations, reach me at z1175085991@gmail.com.

02 / News

Recent updates

  1. Calibration Is Not Control and TAPS were accepted to NeurIPS 2026.

  2. An early version of TAPS was accepted to the ICML 2026 SPIGM workshop.

  3. GoRL and FA-OPD were accepted to ICML 2026.

  4. An early version of GoRL was accepted to the ICLR 2026 ReALM-GEN workshop.

  5. LexChain was accepted to AAAI 2026.

03 / Research

Research

Research diagram comparing risk-score oversight with action-conditioned intervention decisions
NeurIPS 2026

Calibration Is Not Control

Knowing that an agent might fail does not tell us whether intervening will help. We study intervention advantage through action-conditioned decisions and counterfactual branching from the same trajectory prefix.

Preprint arXiv version uses an earlier title
GoRL framework diagram showing latent policy optimization and generative decoder refinement
ICML 2026

Generative Online Reinforcement Learning

GoRL separates tractable latent policy optimization from flow and diffusion action generation, making expressive policies easier to train through online interaction.

Paper Code

04 / Publications

Papers

Conference papers 2026

  1. 01

    Calibration Is Not Control: Intervention Advantage for LLM-Agent Oversight

    Chubin Zhang, Zhenglin Wan, Xingrui Yu, Jingxuan Wu, Qi Wen, Pengfei Zhou, Wangbo Zhao, Ivor Tsang

    NeurIPS 2026

    Preprint ↗
  2. 02

    Denoising Time Matters: Diverse Generation in Diffusion Language Models

    Jingxuan Wu, Zhenglin Wan, Yuzhe Yang, Yiqiao Huang, Chubin Zhang, Xingrui Yu, Ivor Tsang, Yang You

    NeurIPS 2026

    Paper ↗
  3. 03

    Generative Online Reinforcement Learning

    Chubin Zhang, Zhenglin Wan, Feng Chen, Fuchao Yang, Lang Feng, Yaxin Zhou, Xingrui Yu, Yang You, Ivor Tsang, Bo An

    ICML 2026

    Paper ↗Code ↗
  4. 04

    Adversarial Dual On-Policy Distillation from Expressive Teacher

    Zhenglin Wan, Jingxuan Wu, Xingrui Yu, Chubin Zhang, Mingcong Lei, Bo An, Ivor Tsang, Yang You

    ICML 2026

    Paper ↗
  5. 05

    LexChain: Modeling Legal Reasoning Chains for Chinese Tort Case Analysis

    Huiyuan Xie, Chenyang Li, Huining Zhu, Chubin Zhang, Yuxiao Ye, Zhenghao Liu, Zhiyuan Liu

    AAAI 2026

    Paper ↗Code ↗

Workshop papers 2026

  1. 06

    Time-Annealed Perturbation Sampling: Diverse Generation for Diffusion Language Models

    Jingxuan Wu, Zhenglin Wan, Yuzhe Yang, Yiqiao Huang, Chubin Zhang, Xingrui Yu, Ivor Tsang, Yang You

    ICML 2026 Workshop on SPIGM

    Paper ↗
  2. 07

    Decoupling Tilting from Transport: Stable Online Alignment of Flow and Diffusion Policies

    Chubin Zhang, Zhenglin Wan, Feng Chen, Fuchao Yang, Lang Feng, Yaxin Zhou, Xingrui Yu, Yang You, Ivor Tsang, Bo An

    ICLR 2026 Workshop on ReALM-GEN

    Paper ↗

Preprints 2026

  1. 08

    Judged Useless, Queried Anyway: Tool-Using Agents Rarely Turn Their Own Evidence Judgments into Stopping Decisions

    Chubin Zhang, Zhenglin Wan, Xingrui Yu, Jingxuan Wu, Yaxin Zhou, Ivor Tsang, Bo An

    arXiv preprint, 2026

    Preprint ↗Code ↗

05 / Background

Where I have worked

My research background includes online reinforcement learning, generative modeling, and language-model research.

Browse publications
2026 - present

Nanyang Technological University

Research Assistant in Prof. Bo An's group / Singapore

2025 - 2026

A*STAR Centre for Frontier AI Research

Research Intern with Prof. Ivor Tsang / Singapore

2024 - 2025

Tsinghua University, THUNLP

Research Intern with Prof. Zhiyuan Liu / Beijing

06 / Contact

Let's talk research.

For research conversations, collaborations, or opportunities, reach me by email.

z1175085991@gmail.com