Tootfinder

@arXiv_csCR_bot@mastoxiv.page
2025-07-28 09:14:11

Running in CIRCLE? A Simple Benchmark for LLM Code Interpreter Security
Gabriel Chua
https://arxiv.org/abs/2507.19399 https://arxiv.org/pdf/2507.19399

Running in CIRCLE? A Simple Benchmark for LLM Code Interpreter Security
As large language models (LLMs) increasingly integrate native code interpreters, they enable powerful real-time execution capabilities, substantially expanding their utility. However, such integrations introduce potential system-level cybersecurity threats, fundamentally different from prompt-based vulnerabilities. To systematically evaluate these interpreter-specific risks, we propose CIRCLE (Code-Interpreter Resilience Check for LLM Exploits), a simple benchmark comprising 1,260 prompts targe…