This repository contains a model from the Re:Form framework, presented in the paper Re:Form — Reducing Human Priors in Scalable Formal Software Verification with RL in LLMs: A Preliminary Study on Dafny. Re:Form is a novel approach that addresses the challenges of unreliable and unscalable verification processes in Large Language Models (LLMs) that typically operate with informal languages. It proposes grounding LLMs in rigorous formal systems, specifically using the Dafny formal language, to enable automatic and mathematically provable verification of their reasoning processes and outcomes. This capability is pivotal for achieving large-scale, reliable formal software verification. The Re:Form pipeline focuses on reducing human priors in complex programming tasks through an automatic and scalable data curation pipeline. It integrates careful Reinforcement Learning (RL) designs with feedback from a formal language verifier. This methodology allows even small models (e.g., 0.5B) to generate syntactically valid and verifiable Dafny code, surpassing proprietary models and demonstrating stronger generalization to out-of-domain tasks, including on the challenging DafnyComp benchmark.…
Модальности:
Генерация текста
Области применения:
Генерация кода Диалог / чат
Задача: Генерация текста
Автор: Veri-Code
Теги: qwen2, code-generation, formal-verification, dafny, reinforcement-learning, conversational, text-generation-inference, endpoints_compatible
Лайков: 3 | Загрузок: 29
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
