PeopleofSuperintelligence

Researcher

Alvaro Herrasti

Alvaro Herrasti is an AI researcher whose work spans mathematical language processing and embodied AI. He coauthored a tree-transducer approach to Math SAT questions, AI2-THOR and a study of embodied agents learning through gameplay. His research also includes RoboTHOR’s simulation-to-real platform, ProcTHOR’s procedural environments, Holodeck’s prompt-generated scenes, PoliFormer’s robot navigation agent and MolmoSpaces’ robot-policy benchmark. As of October 2026, he is part of the Allen Institute for AI team.

Timeline

  1. 2017

    Math SAT and AI2-THOR

    In 2017, he coauthors a tree-transducer approach to Math SAT questions and AI2-THOR, a framework of 3D indoor scenes where AI agents can navigate and interact with objects.

  2. 2019

    Learning through interactive play

    In 2019, he coauthors a study in which embodied reinforcement-learning agents play Cache, a variant of hide-and-seek, and learn representations encoding object permanence, free space and containment.

  3. 2020

    RoboTHOR and its challenge

    In 2020, he coauthors RoboTHOR, an open simulation-to-real platform for embodied AI, and is among the organizers of the RoboTHOR challenge run by AI2’s PRIOR team.

  4. 2022–January 2023

    ProcTHOR wins at NeurIPS

    In 2022, he coauthors ProcTHOR, a framework for generating diverse virtual environments to train embodied agents; the paper wins the NeurIPS Outstanding Paper Award. By January 2023, he is an AI2 research engineer.

  5. 2023

    Holodeck from language prompts

    In 2023, he coauthors Holodeck, which automatically generates diverse 3D environments from user-supplied prompts for embodied AI agents.

  6. 2024

    PoliFormer for robot navigation

    In 2024, he coauthors PoliFormer, an RGB-only indoor navigation agent trained with large-scale reinforcement learning in simulation that generalizes to the real world without adaptation.

  7. 2026

    MolmoSpaces for robot policies

    In 2026, he coauthors MolmoSpaces, a fully open ecosystem for large-scale robot-policy benchmarking with over 230,000 diverse indoor environments.

  8. As of October 2026

    On the AI2 team

    As of October 2026, he is part of the Allen Institute for AI team.

  1. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Allen Institute for AI (Ai2). 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 Allen Institute for AI (Ai2). 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. 72% of AI researchers with top works connect into a single 2,592-person co-author groupIncludes Allen Institute for AI (Ai2). We map the co-author network across 3,593 people shaping AI and 14,977 most-cited papers: 2,592 (72%) form one connected group with a median distance of 5 steps, led in 2-step reach by Yoshua Bengio (570) and in shortest-path bridging by Yoshua Bengio (6.7%).
  4. Direct Form 990 pay in the latest returns of 39 nonprofits with AI people on this site is highest at $2.1M at the Allen Institute for AI (Ai2) (2024)Includes Allen Institute for AI (Ai2). IRS Form 990 compensation across 39 nonprofits with AI people on this site from 2012 to 2025: in the latest returns, Mozilla reports $6.1M including related-company pay (2024) and the Allen Institute for AI (Ai2) leads direct nonprofit pay at $2.1M (2024).

Sources and credits

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

  1. Beyond Sentential Semantic Parsing: Tackling the Math SAT with a Cascade of Tree Transducers aclanthology.org. Checked 21 October 2026.
  2. AI2-THOR: An Interactive 3D Environment for Visual AI arxiv.org. Checked 21 October 2026.
  3. Learning Generalizable Visual Representations via Interactive Gameplay arxiv.org. Checked 21 October 2026.
  4. RoboTHOR: An Open Simulation-to-Real Embodied AI Platform arxiv.org. Checked 21 October 2026.
  5. RoboTHOR Challenge 2020 ai2thor.allenai.org. Checked 21 October 2026.
  6. ProcTHOR: Large-Scale Embodied AI Using Procedural Generation arxiv.org. Checked 21 October 2026.
  7. Allen School and AI2 researchers recognized at NeurIPS for outstanding contributions in large-scale embodied AI and next-generation image-text models news.cs.washington.edu. Checked 21 October 2026.
  8. AI2 Newsletter | December 2022 allenai.org. Checked 21 October 2026.
  9. Holodeck: Language Guided Generation of 3D Embodied AI Environments arxiv.org. Checked 21 October 2026.
  10. PoliFormer: Scaling On-Policy RL with Transformers Results in Masterful Navigators arxiv.org. Checked 21 October 2026.
  11. MolmoSpaces: A Large-Scale Open Ecosystem for Robot Navigation and Manipulation arxiv.org. Checked 21 October 2026.
  12. Team Directory | Ai2 allenai.org. Checked 21 October 2026.