PeopleofSuperintelligence

Researcher

Baolin Peng

Baolin Peng is a Microsoft Research researcher whose work connects dialogue systems, reinforcement learning, and language-model agents. His early work included ACE for colloquial-Cantonese and written-Chinese error detection, the ConvLab dialogue platforms, and LLM-Augmenter for grounding model responses in external knowledge. He coauthored Magma for multimodal agents and CollabLLM, which received an ICML Outstanding Paper award in 2025. In 2026, he coauthored Orchard, an open-source agentic modeling framework with reusable environments and recipes for software engineering, browser navigation, and assistant tasks. He is a Principal Researcher and Research Manager at Microsoft Research in Redmond and holds a computer science PhD from The Chinese University of Hong Kong.

Timeline

  1. 2016–2017

    ACE and a dialogue policy

    In 2016, he coauthors ACE, which combines n-gram and rule-based models to detect colloquial Cantonese and errors in written Chinese. In 2017, he coauthors a hierarchical deep reinforcement learning method for dialogue managers that select among subtasks and operate across temporal scales.

  2. 2019

    ConvLab's multi-domain platform

    In 2019, he coauthors ConvLab, an open-source multi-domain dialogue platform with reusable components, annotated datasets, and pre-trained reference models for comparing system designs.

  3. 2020

    A toolkit for dialogue systems

    In 2020, he coauthors ConvLab-2, an open-source toolkit for building, evaluating, and diagnosing task-oriented dialogue systems. As ConvLab's successor, it integrates more powerful models and supports more datasets.

  4. 2023

    Grounding language-model responses

    In 2023, he coauthors LLM-Augmenter, which grounds black-box language models in task-specific external knowledge and iteratively revises prompts using feedback to improve responses.

  5. February 2025

    Magma for multimodal agents

    In February 2025, he coauthors Magma, a multimodal foundation model for agents that plans and acts in digital and physical environments, from UI navigation to robot manipulation.

  6. 2025

    An ICML Outstanding Paper

    In 2025, he coauthors CollabLLM, a training framework whose collaborative simulation estimates long-term response value with multiturn-aware rewards and helps language models uncover user intent. ICML gives it an Outstanding Paper award.

  7. 2026

    The Orchard agent framework

    In 2026, he coauthors Orchard, an open-source agentic modeling framework with reusable sandbox environments and recipes for software engineering, browser navigation, and assistant tasks.

  8. As of October 2026

    Principal researcher at Microsoft

    As of October 2026, he is a Principal Researcher and Research Manager at Microsoft Research in Redmond, where he works on AI agents, reasoning and planning, reinforcement learning, and recursive self-improvement. He holds a PhD in computer science from The Chinese University of Hong Kong.

Career

  1. NowMicrosoft ResearchPrincipal Researcher and Research Manager

Education

  1. Chinese University of Hong KongPhD

From the biography.

  1. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Baolin Peng. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.
  2. Among employers with 30+ full-time cases, OpenAI leads primary technical role H-1B pay at $310,000, and Amazon, Google and Microsoft file 57% of all casesIncludes Microsoft. Across 440,669 certified US H-1B filings from 44 employers (FY2020–FY2026), OpenAI leads primary technical role base pay at $310,000 and xAI leads overall base pay at $290,000.
  3. Google had the most people shaping AI every year from 2014 to 2025Includes Microsoft. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  4. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Microsoft. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.

Sources and credits

Photo: We have no published photo of Baolin Peng yet. Send us a free-licensed photo.

  1. ACE: Automatic Colloquialism, Typographical and Orthographic Errors Detection for Chinese Language aclanthology.org. Checked 21 October 2026.
  2. [1704.03084] Composite Task-Completion Dialogue Policy Learning via Hierarchical Deep Reinforcement Learning arxiv.org. Checked 21 October 2026.
  3. [1904.08637] ConvLab: Multi-Domain End-to-End Dialog System Platform arxiv.org. Checked 21 October 2026.
  4. [2002.04793] ConvLab-2: An Open-Source Toolkit for Building, Evaluating, and Diagnosing Dialogue Systems arxiv.org. Checked 21 October 2026.
  5. [2302.12813] Check Your Facts and Try Again: Improving Large Language Models with External Knowledge and Automated Feedback arxiv.org. Checked 21 October 2026.
  6. Magma: A Foundation Model for Multimodal AI Agents - Microsoft Research microsoft.com. Checked 21 October 2026.
  7. [2502.00640] CollabLLM: From Passive Responders to Active Collaborators arxiv.org. Checked 21 October 2026.
  8. ICML 2025 Awards icml.cc. Checked 21 October 2026.
  9. [2605.15040] Orchard: An Open-Source Agentic Modeling Framework arxiv.org. Checked 21 October 2026.
  10. Baolin Peng | AI Agents & Reinforcement Learning pengbaolin.github.io. Checked 21 October 2026.
  11. Baolin Peng at Microsoft Research microsoft.com. Checked 21 October 2026.
  12. People — Orchard-Agentic microsoft.github.io. Checked 21 October 2026.