A new paper on arXiv introduces IneqAgent, an agentic framework designed for formal inequality proving in Lean. According to the abstract, formal inequality proving requires both mathematical reasoning and exact computation, often linked through intermediate lemmas and goal reformulations.

IneqAgent is described as combining LLM reasoning with symbolic computation, suggesting a hybrid strategy in which a language model guides proof search while symbolic tools handle exact arithmetic or algebraic steps. The abstract is truncated, so specific implementation details and experimental results are not available in the source.

Because only the opening of the abstract is provided, the article cannot confirm how IneqAgent compares with prior methods or what benchmarks it was tested on. The reported contribution is the framework itself: an agentic approach that treats inequality proving as a reasoning task supported by exact computation.