Aletheia is a math research agent that generates, verifies, and revises solutions using advanced Gemini Deep Think. It achieves milestones like fully AI-generated papers, human-AI collaborations, and solving four open Erdos problems. The work proposes standards for quantifying AI autonomy in math.
Key Points
- 1.AI-generated paper on eigenweights (Feng26)
- 2.Solved 4 open problems in Erdos database
- 3.Human-AI proof on independent sets (LeeSeo26)
Impact Analysis
Accelerates math research by automating PhD-level tasks and literature navigation. Enables scalable AI-human collaboration. Standardizes evaluation of AI math contributions.
Technical Details
Powered by Gemini Deep Think with inference-time scaling and tool use. Handles long-horizon proofs in natural language. Evaluated on Olympiad to PhD exercises.