- Professor (EHESS), Cognitive machine learning team, EHESS (Laboratoire de Sciences Cognitives et Psycholinguistique)1
Emmanuel Dupoux is a professor at EHESS and a member of the Laboratoire de Sciences Cognitives et Psycholinguistique at the École normale supérieure, in the Cognitive Machine Learning (CoML) team. CoML describes its work as reverse engineering early language acquisition with unsupervised algorithms that discover speech and language units from raw sensory data, and building open-source cognitive tests that compare artificial systems with human performance. Dupoux is a co-author of On Generative Spoken Language Modeling from Raw Audio, a paper whose authors are listed at Facebook AI Research, with Dupoux also at EHESS.1234
At a glance
- Current role
- Professor (EHESS), Cognitive machine learning team
- Company
- EHESS
- Sector
- University
- Work listed
- 1
- Sources cited
- 14
- Profile reviewed
Research
Dupoux leads the Cognitive Machine Learning (CoML) team, whose members page lists Dupoux as an EHESS professor and team leader working on psycholinguistics, cognitive development, speech perception and modeling. The team's main project takes a reverse-engineering approach to infant language learning, building artificial learners trained on recordings of child-adult interaction and testing the resulting algorithms against infant behavior and brain measurements.56
Dupoux described this approach in a 2016 single-author paper proposing a roadmap for reverse-engineering the infant language learner. The same year Dupoux co-authored, with Tal Linzen and Yoav Goldberg, a study of whether LSTM networks learn syntax-sensitive agreement dependencies. In 2026 Dupoux, Yann LeCun and Jitendra Malik proposed a learning architecture inspired by human and animal cognition to address limits on autonomous learning in current AI systems.789
Benchmarks and spoken language models
Dupoux co-authored the papers describing the 2017 and 2020 editions of the Zero Resource Speech Challenge, which asks systems to discover subword and word units from raw speech without labels. The CoML team also maintains a list of open-source tools for unsupervised speech learning, and Dupoux co-authored Libri-Light, a benchmark for speech recognition with limited or no supervision.10111213
Working with researchers at Facebook AI Research, Dupoux co-authored the 2021 paper on generative spoken language modeling from raw audio, and the 2023 TWIST paper, which warm-starts speech language models from pretrained text language models.3414
Work and contributions
Generative Spoken Language Modeling (GSLM) Published research
Last-listed co-author of On Generative Spoken Language Modeling from Raw Audio (2021), which introduces the task of learning a language's acoustic and linguistic characteristics from raw audio without text; the paper lists the authors at Facebook AI Research.34
Timeline
- Posted a roadmap for reverse-engineering the infant language learner as sole author.7
- Co-authored "Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies".8
- Co-authored the paper describing the Zero Resource Speech Challenge 2017.10
- Co-authored Libri-Light, a benchmark for speech recognition with limited or no supervision.13
- Co-authored the paper describing the Zero Resource Speech Challenge 2020.11
- Co-authored "Generative Spoken Language Modeling from Raw Audio" with Facebook AI Research.3
- Co-authored "Textually Pretrained Speech Language Models", introducing the TWIST method.14
- Posted a paper with Yann LeCun and Jitendra Malik on autonomous learning lessons from cognitive science.9
Other documented roles
Roles the sources document that are not confirmed as current.
- Listed with Facebook AI Research, and also at EHESS, on the generative spoken language modeling paper (2021), Facebook AI Research (Meta)4
Where they work
- EHESS 1 person in the wiki
Related people
- Karen LivescuSpeech and audio · Language models
- Nils ReimersLanguage models · Evaluation and benchmarks
- Yulia TsvetkovLanguage models · Evaluation and benchmarks
- Stella BidermanLanguage models · Evaluation and benchmarks
- Oriol VinyalsLanguage models · Speech and audio
- Akshita BhagiaLanguage models · Evaluation and benchmarks
- Kai-Fu LeeSpeech and audio · Language models
- Yacine JerniteLanguage models · Evaluation and benchmarks
- Dan JurafskyLanguage models · Speech and audio
- Eric Michael SmithEvaluation and benchmarks · Language models
Suggest a correction or removal
Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.
References
14 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026
- Emmanuel Dupoux — LSCP member directory
- CoML — Cognitive Machine Learning team overview
- Generative Spoken Language Modeling from Raw Audio (arXiv 2102.01192)
- On Generative Spoken Language Modeling from Raw Audio — paper PDF (author affiliations)
- CoML — People
- CoML — Projects
- Cognitive Science in the era of Artificial Intelligence: A roadmap for reverse-engineering the infant language-learner (arXiv 1607.08723)
- Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies (arXiv 1611.01368)
- Why AI systems don't learn and what to do about it: Lessons on autonomous learning from cognitive science (arXiv 2603.15381)
- The Zero Resource Speech Challenge 2017 (arXiv 1712.04313)
- The Zero Resource Speech Challenge 2020: Discovering discrete subword and word units (arXiv 2010.05967)
- CoML — Resources
- Libri-Light: A Benchmark for ASR with Limited or No Supervision (arXiv 1912.07875)
- Textually Pretrained Speech Language Models (arXiv 2305.13009)