Aletheia tackles FirstProof autonomously

Tony Feng, Junehyuk Jung, Sang-hyun Kim, Carlo Pagano, Sergei Gukov +12 more
2/24/2026
cs.AIcs.CLcs.LG

Official code: google-deepmind/superhuman (766 stars, Apache-2.0, last commit 2026-03-20).

Abstract

We report the performance of Aletheia (Feng et al., 2026b), a mathematics research agent powered by Gemini 3 Deep Think, on the inaugural FirstProof challenge. Within the allowed timeframe of the challenge, Aletheia autonomously solved 6 problems (2, 5, 7, 8, 9, 10) out of 10 according to majority expert assessments; we note that experts were not unanimous on Problem 8 (only). For full transparency, we explain our interpretation of FirstProof and disclose details about our experiments as well as our evaluation. Raw prompts and outputs are available at https://github.com/google-deepmind/superhuman/tree/main/aletheia.

View on arXivView PDF

Code Implementations(1)

google-deepmind/superhumanOfficial100%
Apache-2.0
76679TeXOct 29, 20257 months agoApache-2.0

Similar Papers

Towards Autonomous Mathematics Research

83% match1 repo7662026

Aletheia: Quantifying Cognitive Conviction in Reasoning Models via Regularized Inverse Confusion Matrix

76% match2026

Aristotle: IMO-level Automated Theorem Proving

74% match2025

Accelerating Scientific Research with Gemini: Case Studies and Common Techniques

74% match2026

The Mathematician's Assistant: Integrating AI into Research Practice

72% match2025

Semi-Autonomous Mathematics Discovery with Gemini: A Case Study on the Erdős Problems

72% match2026

Cite this paper

@article{feng2026aletheia,
  title  = {Aletheia tackles FirstProof autonomously},
  author = {Tony Feng and Junehyuk Jung and Sang-hyun Kim and Carlo Pagano and Sergei Gukov and Chiang-Chiang Tsai and David Woodruff and Adel Javanmard and Aryan Mokhtari and Dawsen Hwang and Yuri Chervonyi and Jonathan N. Lee and Garrett Bingham and Trieu H. Trinh and Vahab Mirrokni and Quoc V. Le and Thang Luong},
  year   = {2026},
  eprint = {2602.21201},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2602.21201v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2602.21201.svg)](https://codeofpaper.com/paper/2602.21201)

Discussion