David X. Wu

David X. Wu

he/him/his

PhD Student, UC Berkeley

I am a researcher at OpenAI and a final year Ph.D. student at UC Berkeley, where I'm extremely fortunate to be coadvised by Prasad Raghavendra and Anant Sahai.

I'm grateful to be supported by an NSF GRFP fellowship and an OpenAI Superalignment Grant.

I am interested in LLM reasoning for coding and math. I'm also broadly interested in problems related to theoretical computer science and statistics, such as Markov chains, statistical inference, and machine learning theory.

I graduated with a B.Sc. from MIT with a double major in mathematics (Course 18) and computer science (Course 6-3), where I was lucky to do research with Justin Solomon and Suvrit Sra. Previously, I interned at Windsurf working on coding agents and in quant finance at HRT and Akuna Capital.

Selected Publications and Preprints

  • arXiv Synthetic Error Injection Fails to Elicit Self-Correction In Language Models David X. Wu*, Shreyas Kapur*, Anant Sahai, Stuart Russell arXiv, 2025 arXiv
  • arXiv Markov Chains Approximate Message Passing Amit Rajaraman*, David X. Wu* arXiv, 2025 arXiv
  • STOC Weak Poincaré Inequalities, Simulated Annealing, and Sampling from Spherical Spin Glasses Brice Huang*, Sidhanth Mohanty*, Amit Rajaraman*, David X. Wu* ACM Symposium on Theory of Computing, 2025 arXiv
  • ICLR Provable Weak-to-Strong Generalization via Benign Overfitting David X. Wu, Anant Sahai International Conference on Learning Representations, 2025 arXiv