Reproduce “Silent and Costly: Characterizing and Detecting Prompt-Caching Mis-Integration in LLM Agents” 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
Ying Zhang, Ruoqi Guo, Yi Liu, Gelei Deng, Yuekang Li, Yepang Liu, Leo Zhang
The paper examines prompt-caching integration defects and introduces CACHELINT, a detector for explicit cache markers or parameters. Across 120 defects, 108 raised no error; 97 were request-side and 23 read-side. CACHELINT combines CodeQL candidate-site detection with capability and contextual guard checks. Across 571 repositories, it produced 37 warnings with 0.86 precision; recall is unknown because defects were not exhaustively labeled.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.