Brief

Google DeepMind's AlphaProof Nexus has autonomously solved nine open Erdős mathematical problems, including two that had remained unsolved for 56 years. The AI system, which pairs a large language model with the Lean compiler for automatic proof verification, achieved these breakthroughs at a cost of a few hundred dollars per problem. This development showcases AI's growing capability in generating original mathematical solutions and formal verification. AI

Researchers are exploring the use of agentic AI systems, particularly those leveraging large language models (LLMs), for complex tasks like program verification and mathematical theorem proving. Studies show these systems can achieve high success rates in generating valid specifications and certifying code, sometimes outperforming specialized models on new benchmarks. However, the research also highlights a growing gap between current AI capabilities and the rigor of existing verification benchmarks, suggesting a need for more robust evaluation methods. AI

Researchers have developed Lean Refactor, a new framework designed to optimize proofs generated by large language models (LLMs) in the Lean mathematical proof assistant. This system addresses key challenges such as proof length, compilation cost, and version compatibility, which are often in tension. By using a retrieval-augmented agentic approach with a curated database of refactoring strategies, Lean Refactor achieves significant compression rates and reduces compilation times, outperforming previous methods and demonstrating improved version transfer capabilities. AI