PeopleofSuperintelligence

Researcher

Michel Galley

Michel Galley is a computer scientist whose work connects machine translation, neural dialogue and models for human–AI collaboration. After computer science degrees at EPFL and Columbia and postdoctoral work at Stanford, he coauthored neural conversation research and dialogue systems including DialoGPT and GODEL. In 2024, he coauthored MathVista and a NAACL Outstanding Paper on language-model self-improvement; in 2025, CollabLLM won an ICML Outstanding Paper Award. He is a Senior Principal Research Manager at Microsoft Research, where he leads work on AI models and agents for human–AI collaboration.

Works
165
Citations
14,585
h-index
50
Badges

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. 2001–2002

    Computer science at EPFL and Columbia

    He earns a diploma in computer science at EPFL in 2001 and an M.Sc. in computer science at Columbia in 2002.

  2. 2004–2006

    Syntactic translation papers

    In 2004, he coauthors What’s in a Translation Rule? with Mark Hopkins, Kevin Knight and Daniel Marcu. In 2006, he is first author of Scalable Inference and Training of Context-Rich Syntactic Translation Models, with six coauthors including Jonathan Graehl, Kevin Knight and Daniel Marcu.

  3. 2007–2009

    Columbia Ph.D. and Stanford postdoc

    He completes a computer science Ph.D. at Columbia in 2007 under Kathleen McKeown. From 2007 to 2009, he works as a computer science postdoc at Stanford University under Christopher Manning.

  4. 2015–2017

    Neural dialogue and grounded models

    In 2015, he coauthors A Neural Network Approach to Context-Sensitive Generation of Conversational Responses, described as one of the earliest neural conversation models. In 2016, he coauthors A Diversity-Promoting Objective Function for Neural Conversation Models with Jiwei Li, Chris Brockett, Jianfeng Gao and Bill Dolan. In 2017, he coauthors the preprint A Knowledge-Grounded Neural Conversation Model, which conditions responses on conversation history and external facts.

  5. 2019–2022

    Dialogue models and response ranking

    In 2019, he coauthors DialoGPT, trained on 147 million conversation-like exchanges extracted from Reddit comment chains. In 2020, he coauthors DialogRPT, a set of GPT-2 models trained on 133 million human-feedback pairs to rank responses. In 2022, he coauthors GODEL, which adds grounded pretraining for dialogue tasks that need information outside a conversation; DialoGPT and GODEL become some of the earliest widely used open-weight conversational LLMs.

  6. 2024

    MathVista and a NAACL award

    In 2024, he coauthors MathVista, an ICLR benchmark for mathematical reasoning in visual contexts. He also coauthors Teaching Language Models to Self-Improve through Interactive Demonstrations, which introduces TriPosT to train smaller models to improve their own outputs. The paper wins an Outstanding Paper Award at NAACL 2024.

  7. 2025

    CollabLLM and an ICML award

    In 2025, he coauthors CollabLLM: From Passive Responders to Active Collaborators with Shirley Wu, Baolin Peng, Jianfeng Gao and six other coauthors. In 2025, the framework uses simulated multi-turn interactions and Multiturn-aware Rewards to estimate the long-term contribution of responses. In a 2025 user study with 201 judges, it raises satisfaction by 17.6%, and the paper wins an ICML Outstanding Paper Award.

  8. As of October 2026

    Senior Principal Research Manager

    In July 2025, he serves as a Senior Principal Research Manager at Microsoft Research. As of October 2026, he is a Senior Principal Research Manager in its Deep Learning Group, where he leads a team that builds AI models and agents for human–AI collaboration. His team works on LLM post-training, agentic knowledge discovery and interfaces, user simulation, adaptive UX and AI for education, with CollabLLM among its projects; in September 2026, he coauthors StudentSim, a framework for individualized student simulators that update under tutor guidance.

Career

  1. Since 2014Microsoft ResearchSenior Principal Research Manager
  2. 2022–2024Columbia University
  3. 2017–2022University of Washington
  4. 2011–2026Microsoft
  5. 2007–2009Stanford UniversityPostdoc

Education

  1. From 2002Columbia UniversityM.Sc.; Ph.D.
  2. 2001EPFLDiploma

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

Most-cited works

  1. A Diversity-Promoting Objective Function for Neural Conversation Models2016 · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · 1,978 citationsWith Jianfeng Gao
  2. Deep Reinforcement Learning for Dialogue Generation2016 · Conference on Empirical Methods in Natural Language Processing (EMNLP) · 1,220 citationsWith Dan Jurafsky and Jianfeng Gao
  3. DIALOGPT: Large-Scale Generative Pre-training for Conversational Response Generation2020 · Annual Meeting of the Association for Computational Linguistics: System Demonstrations (ACL Demo) · 1,074 citationsWith Jianfeng Gao
  4. A Persona-Based Neural Conversation Model2016 · Annual Meeting of the Association for Computational Linguistics (ACL) · 897 citations
  5. A Neural Network Approach to Context-Sensitive Generation of Conversational Responses2015 · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · 768 citationsWith Jianfeng Gao and Margaret Mitchell

From OpenAlex, as of 21 October 2026.

  1. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Michel Galley. 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.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

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

  1. Michel Galley at Microsoft Research microsoft.com. Checked 21 October 2026.
  2. What’s in a translation rule? - ACL Anthology aclanthology.org. Checked 21 October 2026.
  3. Scalable Inference and Training of Context-Rich Syntactic Translation Models - ACL Anthology aclanthology.org. Checked 21 October 2026.
  4. A Neural Network Approach to Context-Sensitive Generation of Conversational Responses - ACL Anthology aclanthology.org. Checked 21 October 2026.
  5. A Diversity-Promoting Objective Function for Neural Conversation Models - ACL Anthology aclanthology.org. Checked 21 October 2026.
  6. [1702.01932] A Knowledge-Grounded Neural Conversation Model arxiv.org. Checked 21 October 2026.
  7. DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation - Microsoft Research microsoft.com. Checked 21 October 2026.
  8. Dialogue Response Ranking Training with Large-Scale Human Feedback Data - ACL Anthology aclanthology.org. Checked 21 October 2026.
  9. [2206.11309] GODEL: Large-Scale Pre-Training for Goal-Directed Dialog arxiv.org. Checked 21 October 2026.
  10. MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts proceedings.iclr.cc. Checked 21 October 2026.
  11. Teaching Language Models to Self-Improve through Interactive Demonstrations - ACL Anthology aclanthology.org. Checked 21 October 2026.
  12. CollabLLM: Teaching LLMs to collaborate with users - Microsoft Research microsoft.com. Checked 21 October 2026.
  13. [2502.00640] CollabLLM: From Passive Responders to Active Collaborators arxiv.org. Checked 21 October 2026.
  14. ICML 2025 Awards icml.cc. Checked 21 October 2026.
  15. StudentSim: Training LLM-based Student Simulators - Microsoft Research microsoft.com. Checked 21 October 2026.