New framework ExeCRE boosts LLM code generation reliability

Researchers have developed ExeCRE, a framework designed to improve the reliability of code generated by large language models (LLMs). ExeCRE statistically analyzes execution outputs across numerous random inputs to estimate code reliability, rather than relying on traditional tests or LLM feedback. This method projects execution outputs into consistency signals and uses the Dawid-Skene model to infer latent code reliability. When integrated into self-correction pipelines, ExeCRE significantly reduces misleading feedback, as demonstrated by a drop from 113.2 misleading cases to 14.0 under GPT-5.2 on the LiveCodeBench benchmark. AI

IMPACT Enhances the trustworthiness of LLM-generated code, potentially accelerating adoption in critical applications.

RANK_REASON Academic paper introducing a new method for evaluating LLM-generated code reliability. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Dawid-Skene model
• Gotit.pub
• GPT-5.2
• Hugging Face
• LiveCodeBench
• ScienceCast

• paper
• model release

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