AI Paper Atlas

Research map / Large language models / Agents, code and math

SWE-Perf: Can Language Models Optimize Code Performance on Real-World Repositories?

Xinyi He, Qian Liu, Mingzhe Du, Lin Yan, ZhiJie Fan, Yiming Huang, Yin Zheng, Zejian Yuan, Zejun MA

ICML 2026 · Large language models › Agents, code and math · approach: Software engineering agents · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use OpenHands agent instead of Agentless pipelineClaude-3.7-sonnet · SWE-Perf realisticPerformance: 0.41% -> 2.26%Improved
use repo-level OpenHands instead of single-pass oracle promptingClaude-3.7-sonnet · SWE-PerfPerformance: 1.24% -> 2.26%Improved
use Agentless instead of single-pass oracle promptingClaude-3.7-sonnet · SWE-PerfPerformance: 1.24% -> 0.41%Hurt
compare best LLM method to human expert patchesOpenHands · SWE-PerfPerformance: 2.26% -> 10.85%Hurt
increase the number of target functionsexpert and OpenHands · SWE-Perf realisticPerformanceHurt
increase original runtime of the testexpert vs models · SWE-PerfPerformanceMixed
restrict performance metric to correct examples onlyOpenHands vs alternatives · SWE-PerfPerformance_passImproved
compare OpenHands to expert on sklearnOpenHands · SWE-Perf sklearnPerformance: +0.4%Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy