Dafny as Verification-Aware Intermediate Language for Code Generation

2025-01-10

AI Summary

Ensuring functional correctness of code generated by large language models (LLMs) remains challenging due to the absence of rigorous verification mechanisms in standard generation pipelines. Method: This paper proposes a verification-aware code generation paradigm that employs Dafny as a trusted intermediate representation language. LLMs are guided to produce executable code annotated with formal specifications; a specification-driven compiler then automatically translates the verified Dafny code into target languages (e.g., Python). The entire process operates via natural-language interaction, abstracting away low-level Dafny implementation details from users. Contribution/Results: This work introduces the first deep integration of Dafny into LLM-based code generation, enabling tight coupling between code synthesis and formal verification. Evaluated on the HumanEval benchmark, the approach achieves statistically significant improvements in functional correctness, demonstrating the efficacy and practicality of verification-aware intermediate representations for balancing automation and reliability in LLM-driven software development.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Large Multimodal Models (LMMs)
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Using large language models (LLMs) to generate source code from natural language prompts is a popular and promising idea with a wide range of applications. One of its limitations is that the generated code can be faulty at times, often in a subtle way, despite being presented to the user as correct. In this paper, we explore ways in which formal methods can assist with increasing the quality of code generated by an LLM. Instead of emitting code in a target language directly, we propose that the user guides the LLM to first generate an opaque intermediate representation, in the verification-aware language Dafny, that can be automatically validated for correctness against agreed on specifications. The correct Dafny program is then compiled to the target language and returned to the user. All user-system interactions throughout the procedure occur via natural language; Dafny code is never exposed. We describe our current prototype and report on its performance on the HumanEval Python code generation benchmarks.

Problem

• Large Language Models
• Code Generation
• Error Detection

Innovation

• Super-large Language Models
• Automated Code Verification
• Transparent Code Generation