RLMEval: Evaluating Research-Level Neural Theorem Proving

The introduction of RLMEval marks a significant step forward in evaluating neural theorem proving and proof autoformalization, particularly in the context of research-level mathematics. While large language models have shown promise in controlled settings, their real-world application has been limited. RLMEval aims to bridge this gap by providing a robust evaluation suite that focuses on real-world Lean formalization projects. This development is crucial as it not only enhances the understanding of LLMs' capabilities but also paves the way for more effective applications in complex mathematical reasoning.