Proof Pile 2
Proof Pile 2 is a large dataset of mathematical text and code used to train and benchmark large language models (LLMs) for tasks related to automated theorem proving and mathematical reasoning. Current research focuses on improving LLMs' ability to generate and verify mathematical proofs, extract theorems from existing proofs, and assess the difficulty of mathematical statements, employing techniques like evolutionary algorithms, multimodal learning, and refined search strategies within various formal mathematical languages. This work has implications for improving automated reasoning tools, enhancing mathematical education through automated grading and feedback systems, and advancing the field of artificial intelligence by pushing the boundaries of LLMs' capabilities in complex symbolic reasoning.
Papers
A note on the capacity of the binary perceptron
Dylan J. Altschuler, Konstantin Tikhomirov
Showing Proofs, Assessing Difficulty with GeoGebra Discovery
Zoltán Kovács, Tomás Recio, M. Pilar Vélez
Automated Completion of Statements and Proofs in Synthetic Geometry: an Approach based on Constraint Solving
Salwa Tabet Gonzalez, Predrag Janičić, Julien Narboux