New system Formal Disco generates verified code datasets at scale

Researchers have developed Formal Disco, a scalable system designed to generate large datasets of formally verified programs. This system uses a distributed approach with three types of AI workers: initiators to sketch programs, fixers to resolve verification errors, and extenders to expand existing code. Formal Disco aims to overcome the data scarcity issue in formal verification by creating synthetic data, which has been used to fine-tune open models to match or exceed the performance of Claude Opus 4.5 on verification-relevant tasks. The project also introduces a maximum entropy principle for generating diverse programs and releases significant datasets for Dafny, Verus, and Frama-C. AI

IMPACT Addresses data scarcity in formal verification, potentially accelerating the development and reliability of AI-generated code.

RANK_REASON The cluster describes a research paper detailing a new system and methodology for generating synthetic data for formal verification. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• Claude Opus 4.5
• Dafny
• Formal Disco
• Frama-C
• Hugging Face
• Verus

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →