Search papers, labs, and topics across Lattice.
This paper introduces PerfAgent, a novel workflow that enhances repository-level code optimization for large language model agents by integrating profiler-guided feedback and a verifier-in-the-loop mechanism. By addressing the limitations of existing agents, which often overlook hidden bottlenecks and stop at superficial optimizations, PerfAgent significantly improves the identification of performance hotspots and the quality of code patches. The results demonstrate that PerfAgent doubles the rate of expert-matching patches on challenging benchmarks, achieving 39.2% on GSO and 74% on SWE-fficiency-Lite, while also outperforming a costly oracle baseline.
PerfAgent doubles the rate of expert-level code optimizations by leveraging profiler-guided feedback, revealing hidden performance bottlenecks that traditional methods miss.
Large language model (LLM) agents now perform well on correctness-oriented repository-level tasks, including SWE-Bench issue resolution and feature implementation in real codebases. However, they still struggle with repository-level code optimization, which requires preserving behavior while improving runtime performance. Passing tests is not enough in this setting; a patch must preserve behavior, implement code optimization, and approach expert speedups. Current agents often miss bottlenecks hidden behind abstraction layers and native extensions, stop after shallow speedups, or insufficiently test the code patches that thus may silently break edge cases. We present PerfAgent, a profiler-guided, verifier-in-the-loop workflow that gives an off-the-shelf coding agent the feedback needed to find real hotspots, improve beyond the first passing patch, and use profiler evidence rather than timing alone to decide what to optimize next. On two challenging optimization benchmarks, GSO and SWE-fficiency-Lite, PerfAgent more than doubles the rate of expert-matching patches over OpenHands with GPT-5.1, improving from 19.6% to 39.2% on GSO and from 26% to 74% on SWE-fficiency-Lite. It also surpasses an oracle best-of-five baseline at substantially lower cost, showing that the gains come from better feedback rather than additional test-time sampling.
MIT CSAIL