PeopleofSuperintelligence

Researcher

Emily Dinan

Emily Dinan is a researcher whose career connects mathematics with conversational safety, language-model training and multimodal AI. As a first-year student at the University of Washington in 2015, she coauthored a paper on folding concave polygons; she later earned an MS in mathematics before entering AI research. Her 2019 work tested dialogue safety through adversarial human attacks, and her 2020 paper studied gender bias in generated dialogue. She coauthored 2021 and 2022 work on safer dialogue systems and a 2025 paper on text-guided image representations. By 2026, she had co-led post-training research for Llama 3 and co-created Cicero; as of October 2026, she is a Research Engineer at Meta Superintelligence Labs, focused on multimodal and language modeling.

Links
Google Scholar, LinkedIn and GitHub

Timeline

  1. By 2019

    A paper on L-shaped polygons

    In 2015, while in her first year at the University of Washington, Dinan coauthors a paper on folding concave polygons into convex polyhedra. Before her AI work, she earns a master's degree in mathematics at UW.

  2. 2019

    Adversarial dialogue safety

    In 2019, Dinan coauthors Build it Break it Fix it for Dialogue Safety, which trains dialogue models through an iterative process with humans and models in the loop to resist adversarial attacks.

  3. 2020

    Reducing gender bias in dialogue

    At EMNLP 2020, she coauthors Queens are Powerful too, which applies counterfactual data augmentation, targeted data collection and bias-controlled training to reduce gender bias in dialogue.

  4. 2021

    Bot-Adversarial Dialogue

    In 2021, she coauthors Bot-Adversarial Dialogue, which uses human-and-model-in-the-loop evaluation to test chatbot toxicity and explores two methods for safer conversational agents.

  5. 2022

    SafetyKit for chatbot safety

    In 2022, she coauthors SafetyKit, which surveys safety risks in end-to-end conversational AI and introduces a taxonomy of the Instigator, Yea-Sayer and Impostor effects.

  6. November 2025

    A text-guided image encoder

    In November 2025, she coauthors Text-Guided Semantic Image Encoder, which generates image representations conditioned on the input text query.

  7. By 2026

    Llama 3 post-training

    By 2026, she has co-led post-training research for Llama 3 and co-created Cicero, the first AI capable of playing Diplomacy at a human level.

  8. As of October 2026

    Meta Superintelligence Labs

    As of October 2026, she is a Research Engineer at Meta Superintelligence Labs in New York, where her research spans multimodal and language modeling and currently focuses on mixed-modal modeling.

  1. Google had the most people shaping AI every year from 2014 to 2025Our estimates of which organizations employed the most people shaping AI from 1961 to 2025, with a year-by-year chart.
  2. The United States employs an estimated 53% of people in our data with a known employment countryOur estimates of where AI careers are concentrated now and how countries’ shares of dated positions changed over time.
  3. 46% of living people in our data with known birth and employment countries work outside their birth countryOur estimate of cross-border careers in AI, from country of birth and the country of each current organization.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionOur estimates of where the people shaping AI studied and where their careers led, from education records in cited biographies.

Sources and credits

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

  1. Folding concave polygons into convex polyhedra: The L-Shape scholar.rose-hulman.edu. Checked 19 October 2026.
  2. Communicating Employment Opportunities – Association for Women in Mathematics (AWM) awm-math.org. Checked 19 October 2026.
  3. Build it Break it Fix it for Dialogue Safety: Robustness from Adversarial Human Attack arxiv.org. Checked 19 October 2026.
  4. Queens are Powerful too: Mitigating Gender Bias in Dialogue Generation aclanthology.org. Checked 19 October 2026.
  5. Bot-Adversarial Dialogue for Safe Conversational Agents aclanthology.org. Checked 19 October 2026.
  6. SafetyKit: First Aid for Measuring Safety in Open-domain Conversational Systems aclanthology.org. Checked 19 October 2026.
  7. Text-Guided Semantic Image Encoder arxiv.org. Checked 19 October 2026.