Tag
Presents FormalAnalyticGeo, a neural-symbolic framework for automatic generation of multimodal analytic geometry problems using a formal intermediate representation (CDL) and LLM components, yielding the AnalyticGeo7K dataset of over 7K verified problems with high geometric precision.
A technical report introducing the APLR(1) algorithm for generating compact LR(1) parsers, which is simpler and more capable than the existing IELR(1) algorithm, and supports nondeterministic/ambiguous grammars for GLR parsing.
A deep dive into the limitations of regex for parsing HTML, inspired by the famous Stack Overflow answer, discussing formal language theory and the power of industrial regex engines.
Introduces Lean4Agent, a framework using Lean4 for formal modeling and verification of agent workflows and trajectories, demonstrating improved performance on SWE-Bench and ELAIP-Bench.