PeopleofSuperintelligence

Researcher

DiJia (Andy) Su

Also known as Andy Su

DiJia (Andy) Su is an AI researcher whose work spans reinforcement learning, motion forecasting, search planning, and language-model reasoning. He earned a Princeton PhD in machine learning in 2022 and coauthored research on scene-centric motion forecasting. His papers used Transformers for Sokoban planning and explored fast and slow reasoning with latent-token methods. As of October 2026, he is a senior staff research scientist at Meta Superintelligence Labs, where he works on Muse Spark and leads an effort in agentic LLM post-training and reinforcement learning.

Works
16
Citations
61
h-index
4
Badges

Publication numbers from OpenAlex, as of 19 October 2026.

Timeline

  1. 2020

    ConQUR for deep Q-learning

    At ICML 2020, he coauthors ConQUR, which develops methods to mitigate delusional bias in deep Q-learning.

  2. 2022

    A Princeton PhD and motion forecasting

    He earns a Princeton PhD in 2022 with a thesis titled Topics in Machine Learning. In 2022, he coauthors a motion-forecasting paper that uses knowledge distillation to narrow the gap between agent-centric and scene-centric models.

  3. 2024

    Planning with Searchformer

    In 2024, he coauthors Beyond A*, which trains an encoder-decoder Transformer to predict A* search dynamics. Searchformer solves 93.7% of previously unseen Sokoban puzzles in the paper’s evaluation.

  4. 2025

    Dualformer and Token Assorted

    At ICLR 2025, he coauthors Dualformer, a Transformer that integrates fast and slow reasoning modes through randomized reasoning traces. He also coauthors Token Assorted, which mixes latent and text tokens for language-model reasoning.

  5. As of October 2026

    Senior scientist at Meta

    As of October 2026, he is a senior staff research scientist at Meta Superintelligence Labs, where he works on Muse Spark and leads an effort in agentic LLM post-training and reinforcement learning.

Career

  1. NowMeta Superintelligence LabsSenior Staff Research Scientist

Education

  1. 2022Princeton UniversityPhD

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

Most-cited works

  1. Reasoning Beyond Words ? Exploring framework for hidden state reasoning2024 · 18 citationsWith Jason Weston
  2. Narrowing the coordinate-frame gap in behavior prediction models: Distillation for efficient and accurate scene-centric motion forecasting2022 · IEEE International Conference on Robotics and Automation · 13 citations
  3. Autonomous robot navigation with self-learning for collision avoidance with randomly moving obstacles2014 · International Conference on Computer Science & Education (ICCSE) · 5 citations
  4. MUSBO: Model-based Uncertainty Regularized and Sample Efficient Batch Optimization for Deployment Constrained Reinforcement Learning2021 · 5 citations
  5. ConQUR: Mitigating Delusional Bias in Deep Q-learning2020 · 4 citationsWith Craig Boutilier

From OpenAlex, as of 19 October 2026.

  1. The median h-index is 53 at universities and 19 at frontier AI labsIncludes DiJia (Andy) Su. The median h-index is 53 at universities and 19 at frontier AI labs. We compare publishing careers, citations and moves to estimate how universities and frontier labs contribute to AI research.
  2. 14% of AI career moves in the 2020s went from a university to a companyIncludes DiJia (Andy) Su. Our estimates of how university-to-company career transitions changed by decade, including roles added alongside an earlier job.
  3. Google alumni founded or led 87 other organizations, the most among the labs we compareIncludes Meta. We trace 212 organizations founded or led by AI lab alumni, and 223 career moves between the labs.
  4. The company share of AI researchers’ top papers rose from 2% to 32%Includes Meta. We estimate where leading AI researchers worked when their most-cited papers appeared. The company share rose from 2% in 1975–1979 to 32% in 2020–2024.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of DiJia (Andy) Su yet. Send us a free-licensed photo.

  1. ConQUR: Mitigating Delusional Bias in Deep Q-Learning proceedings.mlr.press. Checked 19 October 2026.
  2. DiJia Su csauthors.net. Checked 19 October 2026.
  3. DiJia (Andy) Su andysu.org. Checked 19 October 2026.
  4. [2206.03970] Narrowing the Coordinate-frame Gap in Behavior Prediction Models: Distillation for Efficient and Accurate Scene-centric Motion Forecasting arxiv.org. Checked 19 October 2026.
  5. [2402.14083] Beyond A*: Better Planning with Transformers via Search Dynamics Bootstrapping arxiv.org. Checked 19 October 2026.
  6. Dualformer: Controllable Fast and Slow Thinking by Learning with Randomized Reasoning Traces proceedings.iclr.cc. Checked 19 October 2026.