PulseAugur
EN
LIVE 09:58:35

New benchmark WeSCE measures security drift in LLM-edited code

By PulseAugur Editorial · [1 sources] ·

Researchers have developed WeSCE, a new benchmark designed to measure security drift in code edited by large language models (LLMs). This benchmark includes 400 executable programs derived from real-world code, covering various editing tasks like bug fixing and refactoring. WeSCE proposes a continuous risk representation to aggregate vulnerability signals and measures changes in overall risk, worst-case severity, and vulnerability distribution. AI

IMPACT This benchmark will help researchers and developers better understand and mitigate security risks introduced by LLMs in code generation and editing.

RANK_REASON The cluster contains a research paper detailing a new benchmark for LLM-driven code editing. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

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

New benchmark WeSCE measures security drift in LLM-edited code

COVERAGE [1]

  1. arXiv cs.AI TIER_1 English(EN) · Zhiyu Zhang, Tingyue Wen, Senke Sun, Dengxiang Liang, Enhao Huang ·

    WeSCE: A Benchmark for Measuring Security Drift in LLM-Driven Code Editing

    arXiv:2608.15092v1 Announce Type: cross Abstract: In this work, we introduce WeSCE, a benchmark for quantifying security drift in code editing under weak-security constraints, where tasks specify only functional objectives without explicit security requirements. WeSCE consists of…

RELATED ENTITIES

RELATED TOPICS