PeopleofSuperintelligence

Researcher

Marie-Anne Lachaux

Marie-Anne Lachaux is an AI researcher whose work connects code translation, machine translation, theorem proving and language models. She coauthored studies on unsupervised code translation and diverse neural translation, the DOBF code objective, and HyperTree Proof Search. Her model papers include LLaMA, Mistral 7B and Mixtral. Her career includes Meta in Paris and Mistral AI, where she worked as an AI Research Engineer and founding team member. Since November 2025, she is an independent GenAI expert advising startup teams.

Works
14
Citations
8,508
h-index
13
Badges

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. 2020

    Code and machine translation

    In 2020, she coauthors Unsupervised Translation of Programming Languages with Baptiste Roziere, Lowik Chanussot and Guillaume Lample; their transcompiler translates functions between C++, Java and Python using monolingual source code. She also coauthors Target Conditioning for One-to-Many Generation with Armand Joulin and Guillaume Lample in 2020; it conditions a neural machine translation decoder on a target-domain variable to generate diverse translations.

  2. 2021

    DOBF code pretraining

    In 2021, she coauthors DOBF, a pretraining objective that uses programming-language structure to train a model to recover obfuscated source code.

  3. 2022

    HyperTree Proof Search

    In 2022, she coauthors HyperTree Proof Search, an online-trained transformer prover that learns from earlier proof searches. Online training on unproved Metamath theorems raises accuracy to 82.6%.

  4. 2023

    LLaMA foundation models

    In 2023, she coauthors LLaMA: Open and Efficient Foundation Language Models, a family of foundation models ranging from 7B to 65B parameters trained on publicly available datasets.

  5. 2023

    Mistral 7B

    She coauthors Mistral 7B in 2023, a 7-billion-parameter language model that outperforms Llama 2 13B on all evaluated benchmarks.

  6. 2024

    Mixtral of Experts

    In 2024, she coauthors Mixtral of Experts, a sparse mixture-of-experts language model that uses 13 billion active parameters during inference. It matches or outperforms Llama 2 70B and GPT-3.5 across evaluated benchmarks.

  7. By 2025

    From Meta to Mistral AI

    By 2025, she holds an engineering degree from Télécom Paris and an MSc in computer graphics, vision and imaging from University College London. She passes through Meta in Paris before joining Mistral AI as an AI Research Engineer and a founding team member.

  8. Since November 2025

    Independent GenAI consulting

    In November 2025, she launches her own GenAI activity as an independent expert, auditing and advising technical teams at startups.

Career

  1. 2019–2021Meta
  2. Mistral AIAI Research Engineer, Founding team member

Education

  1. Télécom ParisEngineering degree
  2. University College LondonMSc in computer graphics, vision and imaging

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

Most-cited works

  1. LLaMA: Open and Efficient Foundation Language Models2023 · 4,523 citations
  2. Behavioral Safety and Context Retention of Large Language Models in a Longitudinal ICU Simulation under Offline Conditions2023 · 2,598 citationsWith Adina Williams, Hakan Inan, Sergey Edunov and Sharan Narang
  3. Mistral 7B2023 · 323 citationsWith Alexandre Sablayrolles, Chris Bamford, Diego de las Casas and Teven Le Scao
  4. CCNet: Extracting High Quality Monolingual Datasets from Web Crawl Data2019 · 252 citations
  5. Poly-encoders: Architectures and Pre-training Strategies for Fast and Accurate Multi-sentence Scoring2020 · International Conference on Learning Representations · 171 citationsWith Jason Weston

From OpenAlex, as of 21 October 2026.

  1. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Meta. 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.
  2. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Meta. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.
  3. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Meta. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.
  4. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Meta. Our estimates of where OpenAI recruits came from and where 133 people with former OpenAI roles work now, including the companies and labs they founded.

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 Marie-Anne Lachaux yet. Send us a free-licensed photo.

  1. Unsupervised Translation of Programming Languages arxiv.org. Checked 21 October 2026.
  2. Target Conditioning for One-to-Many Generation aclanthology.org. Checked 21 October 2026.
  3. DOBF: A Deobfuscation Pre-Training Objective for Programming Languages proceedings.neurips.cc. Checked 21 October 2026.
  4. HyperTree Proof Search for Neural Theorem Proving papers.nips.cc. Checked 21 October 2026.
  5. LLaMA: Open and Efficient Foundation Language Models arxiv.org. Checked 21 October 2026.
  6. Mistral 7B arxiv.org. Checked 21 October 2026.
  7. Mixtral of Experts arxiv.org. Checked 21 October 2026.
  8. Ces six femmes ingénieures en IA qui changent le visage de la tech française (La Tribune) telecom-paris.fr. Checked 21 October 2026.