Reproduce “Trim My View: An LLM-Based Code Query System for Module Retrieval in Robotic Firmware” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sima Arasteh, Pegah Jandaghi, Nicolaas Weideman, Dennis Perepech, Mukund Raghothaman, Christophe Hauser, Luis Garcia
USC and partner institutions introduce ChatCPS, a novel code query system that enables efficient module-level analysis of robotic firmware through innovative combination of binary decomposition and open-source LLMs, achieving 0.68 F1-score in identifying critical functionality while working directly with stripped binaries through a cost-effective approach.
Key insights
- Module-level analysis provides better understanding of functionality compared to individual function analysis
- Open-source LLMs can effectively summarize and categorize decompiled code when properly prompted
- Different LLMs show varying trade-offs between accuracy and processing speed
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.