PeopleofSuperintelligence

Researcher

Nouha Dziri

Nouha Dziri is an AI researcher whose work examines the reasoning limits and safety of large language models. After co-creating the FaithDial benchmark for hallucination-free dialogue in 2022, she earned a PhD at the University of Alberta in 2023 under Osmar Zaiane, and was first author of the NeurIPS 2023 paper "Faith and Fate" on the limits of transformers in compositional tasks. At the Allen Institute for AI, she worked with Yejin Choi, announced the Ai2 Safety Toolkit in 2024, and co-led safety and post-training for the OLMo models. The 2025 paper "Artificial Hivemind", which she co-authored, won a NeurIPS Best Paper Award for showing that different language models give strikingly similar answers. At Cohere Labs, she is a senior research scientist who works with Joelle Pineau on multi-agent LLM systems.

Now at
Cohere
Badges

Timeline

  1. By 2022

    The FaithDial benchmark

    As a PhD candidate at the University of Alberta, she works within the Alberta Machine Intelligence Institute under Osmar Zaiane on trustworthy conversational models. In 2022, she and her co-authors, including Siva Reddy, publish FaithDial in Transactions of the Association for Computational Linguistics, a benchmark of hallucination-free dialogues made by editing hallucinated responses in the Wizard of Wikipedia benchmark.

  2. 2023

    A PhD on faithful dialogue

    She earns her PhD from the University of Alberta and Amii in 2023, with a thesis titled "Towards Building Coherent and Faithful Conversational Models" advised by Osmar Zaiane in Computing Science. The thesis addresses the hallucinations, factually incorrect statements, that dialogue systems powered by large pre-trained language models can produce.

  3. 2023

    Faith and Fate on compositionality

    In 2023, she is first author of the NeurIPS paper "Faith and Fate: Limits of Transformers on Compositionality", written with Yejin Choi and other co-authors. It tests transformer LLMs on multi-digit multiplication, logic grid puzzles and a dynamic programming problem, and its findings suggest that these models reduce multi-step compositional reasoning to linearized subgraph matching.

  4. 2024

    The Ai2 Safety Toolkit

    In June 2024, she announces the Ai2 Safety Toolkit, a hub for open collaboration on LLM safety whose first two parts are the WildTeaming red-teaming framework and the WildGuard moderation tool. WildTeaming enables the 262K-example WildJailbreak safety training dataset, and WildGuard, a NeurIPS 2024 paper she co-authors, comes with WildGuardMix, a 92K-example moderation dataset covering 13 risk categories.

  5. By 2025

    Safety lead and OLMo post-training

    By 2025, she is Safety Lead at the Allen Institute for AI and a postdoctoral researcher working with Yejin Choi and the Mosaic team. She co-leads the safety and post-training effort for the OLMo models and leads the creation of a safety suite for OLMo.

  6. 2025

    A NeurIPS Best Paper Award

    In 2025, she is among the authors of "Artificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond)", which receives a NeurIPS 2025 Best Paper Award. The paper introduces Infinity-Chat, a dataset of 26K open-ended user queries, and reports that different language models produce strikingly similar outputs.

  7. As of October 2026

    Senior researcher at Cohere Labs

    She is a senior research scientist at Cohere Labs, working with Joelle Pineau on advancing the science of multi-agent LLM systems, after her time as a research scientist at Ai2. She is a keynote speaker at ODSC West 2026, with a talk titled "Scaling Inference-Time Compute with Teams of Agents for Software Engineering".

Career

  1. NowCohereSenior research scientist
  2. AmiiResearcher
  3. Allen Institute for AI (Ai2)Safety Lead; postdoctoral researcher

Education

  1. 2023University of AlbertaPhD

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 Nouha Dziri. 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. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Cohere. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.
  3. 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 Cohere. 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.
  4. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Cohere. 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.

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 Nouha Dziri yet. Send us a free-licensed photo.

  1. The Stanford Natural Language Processing Group nlp.stanford.edu. Checked 22 October 2026.
  2. FaithDial: A Faithful Benchmark for Information-Seeking Dialogue - ACL Anthology aclanthology.org. Checked 22 October 2026.
  3. Towards Building Coherent and Faithful Conversational Models ualberta.scholaris.ca. Checked 22 October 2026.
  4. ODSC AI West Speakers 2026 - Open Data Science Conference odsc.ai. Checked 22 October 2026.
  5. Faith and Fate: Limits of Transformers on Compositionality proceedings.neurips.cc. Checked 22 October 2026.
  6. The Ai2 Safety Toolkit: Datasets and models for safe and responsible LLMs development | Ai2 allenai.org. Checked 22 October 2026.
  7. NeurIPS Poster WildGuard: Open One-stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs neurips.cc. Checked 22 October 2026.
  8. Nouha DZIRI | Singapore Conference on AI scai.gov.sg. Checked 22 October 2026.
  9. Nouha Dziri nouhadziri.github.io. Checked 22 October 2026.
  10. Announcing the NeurIPS 2025 Best Paper Awards – NeurIPS Blog blog.neurips.cc. Checked 22 October 2026.