My research interests lie at the interface of artificial intelligence, theoretical computer science, and information theory. I am currently seeking conceptual insights into how contemporary AI systems learn through the lens of computational learning theory.

I am fortunate to be advised by Nathan Srebro.

Some Papers

• Learning to Think from Multiple Thinkers Nirmit Joshi, Roey Magen, Nathan Srebro, Nikolaos Tsilivis, Gal Vardi

  Preprint

• Learning to Answer from Correct Demonstrations Nirmit Joshi, Gene Li, Siddharth Bhandari, Shiva Kasiviswanathan, Cong Ma, Nathan Srebro

  ICLR 2026

• Learning single-index models via harmonic decomposition Nirmit Joshi, Hugo Koubbi, Theodor Misiakiewicz, Nathan Srebro

  NeurIPS 2025

• A Theory of Learning with Autoregressive Chain of Thought Nirmit Joshi, Gal Vardi, Adam Block, Surbhi Goel, Zhiyuan Li, Theodor Misiakiewicz, Nathan Srebro

  COLT 2025

• Community Detection in the Hypergraph SBM: Exact Recovery Given the Similarity Matrix Julia Gaudio, Nirmit Joshi

  COLT 2023