PulseAugur
EN
LIVE 09:35:06

New framework Code-MUE measures uncertainty in code LLMs

By PulseAugur Editorial · [3 sources] ·

Researchers have developed Code-MUE, a novel framework designed to measure the uncertainty of code Large Language Models (LLMs). This purely black-box system utilizes execution-based Semantic Interaction Graphs to assess uncertainty by analyzing runtime behavior and calculating the Von Neumann entropy of the solution space. Empirical studies involving eight state-of-the-art LLMs show that Code-MUE effectively correlates with functional correctness, outperforming traditional lexical and embedding-based methods for risk detection in software engineering workflows. AI

IMPACT This framework could improve the reliability and safety of code generation tools by better quantifying model uncertainty.

RANK_REASON The cluster describes a new research paper introducing a novel framework for evaluating code LLMs.

Read on arXiv cs.CL →

AI-generated summary · Google Gemini · from 3 sources. How we write summaries →

New framework Code-MUE measures uncertainty in code LLMs

COVERAGE [3]

  1. arXiv cs.AI TIER_1 English(EN) · Xiaoning Ren, Yinxing Xue, Lei Ma, Yuheng Huang ·

    Code-MUE: Measuring Code LLMs' Uncertainty through Execution-based Semantic Interaction Graphs

    arXiv:2607.12273v1 Announce Type: cross Abstract: As Code Large Language Models (LLMs) become central to modern software engineering, their inherent stochasticity poses significant real-world risks, where even minor errors can lead to severe functional, security, or safety conseq…

  2. arXiv cs.CL TIER_1 English(EN) · Yuheng Huang ·

    Code-MUE: Measuring Code LLMs' Uncertainty through Execution-based Semantic Interaction Graphs

    As Code Large Language Models (LLMs) become central to modern software engineering, their inherent stochasticity poses significant real-world risks, where even minor errors can lead to severe functional, security, or safety consequences. Reliable automation, therefore, demands th…

  3. Hugging Face Daily Papers TIER_1 English(EN) ·

    Code-MUE: Measuring Code LLMs' Uncertainty through Execution-based Semantic Interaction Graphs

    As Code Large Language Models (LLMs) become central to modern software engineering, their inherent stochasticity poses significant real-world risks, where even minor errors can lead to severe functional, security, or safety consequences. Reliable automation, therefore, demands th…

RELATED ENTITIES

RELATED TOPICS