New benchmark MINIF2F-DAFNY tests LLMs for mathematical theorem proving

Researchers have developed MINIF2F-DAFNY, a new benchmark for evaluating Large Language Models (LLMs) in mathematical theorem proving. This system translates the miniF2F benchmark to Dafny, an auto-active verifier, enabling LLMs to guide proof generation while Dafny's automated theorem prover handles low-level details. In evaluations, the best-performing LLM, Claude Opus-4.6, achieved a 62.7% cumulative pass rate, significantly improving upon the baseline performance. AI

IMPACT This benchmark could accelerate the development of LLMs capable of complex mathematical reasoning and formal verification.

RANK_REASON The cluster describes a new benchmark and evaluation for LLMs in mathematical theorem proving, published on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Claude Opus-4.6
• Dafny
• Instituto Todos Pela Saúde
• Mantas Baksys
• miniF2F
• MINIF2F-DAFNY
• satisfiability modulo theories

• paper
• model release

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