Oct 1, 2023 · In this work, we present LEGO-Prover, which employs a growing skill library containing verified lemmas as skills to augment the capability of ...

LEGO-Prover utilizes LLMs to retrieve skills from the growing library and decompose overall informal theorems into small snippets in a step-by-step style.

LEGO-Prover contains a prover and an evolver, which are bridged by the growing skill library. The prover takes the problem's formal statement as input and ...

Registration Required. You must be logged in to view this content. Successful Page Load. ICLR uses cookies to remember that you are logged in.

Lego-prover: Neural theorem proving with growing libraries. H Xin, H Wang, C ... MUSTARD: Mastering Uniform Synthesis of Theorem and Proof Data. Y Huang, X Lin ...

Oct 1, 2023 · During the proving process, LEGO-Prover also manages to generate over 20,000 skills (theorems/lemmas) and adds them to the growing library. Our ...

Figure 1 for LEGO-Prover: Neural Theorem Proving with Growing Libraries. Despite the success of large language models (LLMs), the task of theorem proving ...

Using this data, we develop ReProver (Retrieval-Augmented Prover): an LLM-based prover augmented with retrieval for selecting premises from a vast math library.

Title: LEGO-Prover: Neural Theorem Proving with Growing Libraries. Authors: Haiming Wang, Huajian Xin, Chuanyang Zheng, Lin Li, Zhengying Liu, Qingxing Cao ...

The current state-of-the-art on miniF2F-test is Thor + expert iteration on autoformalised theorems. See a full comparison of 23 papers with code.