Hugging Face Daily Papers · August 11, 2026 · 4 min read

SWE-Bench ProMax: Benchmarking Agents on Large-Scale Multilingual Code Refactoring

Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.

Like Read original ↗

New benchmark for coding agents with more complex problems.</p>\n","updatedAt":"2026-08-11T02:46:52.217Z","author":{"_id":"645b0c3ec35da9c7afd95421","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/645b0c3ec35da9c7afd95421/vYBrCDagHsXAo6J2p-uG0.jpeg","fullname":"Yuling","name":"YerbaPage","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":110,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8621107339859009},"editors":["YerbaPage"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/645b0c3ec35da9c7afd95421/vYBrCDagHsXAo6J2p-uG0.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.09802","authors":[{"_id":"6a7a8cd1019ce76dc7b3aa3e","name":"Yuling Shi","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa3f","name":"Jinghan Xu","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa40","name":"Kelin Fu","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa41","name":"Wenhao Zeng","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa42","name":"Shilin He","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa43","name":"Lei Zhang","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa44","name":"Yue Liu","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa45","name":"Zelin Zhao","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa46","name":"Terry Yue Zhuo","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa47","name":"Jialun Cao","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa48","name":"Siyu Ye","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa49","name":"Tianyu Liu","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa4a","name":"Kai Cai","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa4b","name":"Shing-Chi Cheung","hidden":false},{"_id":"6a7a8cd1019ce76dc7b3aa4c","name":"Xiaodong Gu","hidden":false}],"publishedAt":"2026-08-10T00:00:00.000Z","submittedOnDailyAt":"2026-08-11T00:00:00.000Z","title":"SWE-Bench ProMax: Benchmarking Agents on Large-Scale Multilingual Code Refactoring","submittedOnDailyBy":{"_id":"645b0c3ec35da9c7afd95421","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/645b0c3ec35da9c7afd95421/vYBrCDagHsXAo6J2p-uG0.jpeg","isPro":false,"fullname":"Yuling","user":"YerbaPage","type":"user","name":"YerbaPage"},"summary":"As AI coding agents take on increasingly complex, long-horizon software engineering tasks, existing benchmarks are rapidly saturating and their evaluation quality has come under serious scrutiny: a recent audit found that nearly 60% of unsolved SWE-bench Verified instances contain flawed tests -- either overly narrow tests that reject correct solutions or overly broad tests that check unstated requirements -- and that frontier models can verbatim reproduce gold patches from training data. Code refactoring, which requires coordinated, behavior-preserving changes across many files, offers a substantially harder and more realistic test of agent capability, yet remains underserved by current benchmarks. We introduce SWE-Bench ProMax, an expert-curated, multilingual code refactoring benchmark of 170 instances drawn from real commits across seven programming languages (Python, Java, TypeScript, Go, C, C++, and Rust). Every instance undergoes rigorous, multi-stage curation that directly addresses the quality problems identified in prior benchmarks: issue descriptions are rewritten from scratch to provide precise, unambiguous specifications, and test suites are manually reviewed to remove overly narrow and overly broad tests. Tasks with insufficient complexity or limited cross-file scope are filtered out, yielding a benchmark of challenging, large-scale refactoring tasks that average 11.4 modified files and 261.6 lines of code per instance, substantially exceeding the scale of existing benchmarks. Experiments with frontier models under two agent scaffolds show that the best model achieves only 41.2% resolve rate, confirming that SWE-Bench ProMax presents a meaningful and unsaturated challenge for current AI coding agents. Our benchmark is available at https://huggingface.co/datasets/swe-bench-promax/SWE-Bench-ProMax.","upvotes":115,"discussionId":"6a7a8cd1019ce76dc7b3aa4d","projectPage":"https://huggingface.co/datasets/swe-bench-promax/SWE-Bench-ProMax","ai_summary":"SWE-Bench ProMax is a rigorously curated multilingual benchmark of large-scale code refactoring tasks that reveals substantial unsolved challenges for current AI coding agents.","ai_keywords":["SWE-Bench ProMax","code refactoring","multilingual benchmark","cross-file refactoring","agent scaffolds","resolve rate"],"ai_summary_model":"thinkingmachines/Inkling-Small","organization":{"_id":"653b817d32c97d0655575872","name":"ByteDance","fullname":"ByteDance","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6535c9e88bde2fae19b6fb25/0clr54wj5Ly-RkYU9OXPp.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"64b75ba000bac1088cea9231","avatarUrl":"/avatars/5a8bb7cccfe20dad3fda4405fb1eb3ff.svg","isPro":false,"fullname":"Siyu Ye","user":"SiyuYe1","type":"user"},{"_id":"645b0c3ec35da9c7afd95421","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/645b0c3ec35da9c7afd95421/vYBrCDagHsXAo6J2p-uG0.jpeg","isPro":false,"fullname":"Yuling","user":"YerbaPage","type":"user"},{"_id":"635b482862fb2bc1b52d6be3","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1666926580685-noauth.png","isPro":false,"fullname":"FuCao","user":"ythere","type":"user"},{"_id":"68e3580e71512dfe661ff949","avatarUrl":"/avatars/5355152675c51dca4988ac0c878d47d9.svg","isPro":false,"fullname":"schwa","user":"schwa2003","type":"user"},{"_id":"65684c80a9a1a6a50d779f58","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/65684c80a9a1a6a50d779f58/it534ZdH5LxRub1M_o3uM.jpeg","isPro":false,"fullname":"Silin Chen","user":"Silin-Chen","type":"user"},{"_id":"69770caa2810079eaf632ab1","avatarUrl":"/avatars/0d4099ea4eb0e382e33c075330851cb7.svg","isPro":false,"fullname":"M","user":"Resfeber226","type":"user"},{"_id":"66b9fe571caaa1d77c44e79e","avatarUrl":"/avatars/19dc7c2b8ad9ad3de65fa2b8271be2eb.svg","isPro":false,"fullname":"Yiwen Zhang","user":"yiwenz","type":"user"},{"_id":"68e2570e6ecc8f79ab4577ec","avatarUrl":"/avatars/408e8683b6947b1315c9eb41da5b4a34.svg","isPro":false,"fullname":"aa","user":"Disaaad","type":"user"},{"_id":"6787e230e6eadad8572ae070","avatarUrl":"/avatars/c02eddeb7efcb239aef27bd93e6295f3.svg","isPro":false,"fullname":"Miyazaki","user":"miiyazaki","type":"user"},{"_id":"65e933fcacb29a596cd6eb6e","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/65e933fcacb29a596cd6eb6e/dlH14rO2I4EUMVyjEjdRK.jpeg","isPro":false,"fullname":"Xiaobo Wang","user":"Yofuria","type":"user"},{"_id":"67b0a66ed6e6e0b31fd9196e","avatarUrl":"/avatars/7286b249f7d0a7c261cf64b39bf6c2c1.svg","isPro":false,"fullname":"Takumi","user":"TaNakamura","type":"user"},{"_id":"67b1910494bbfd169134f568","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/67b1910494bbfd169134f568/2iLw4n2qe2fpHFY0eenyT.png","isPro":false,"fullname":"ChaiwenJ","user":"Puggyoll","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":3,"organization":{"_id":"653b817d32c97d0655575872","name":"ByteDance","fullname":"ByteDance","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6535c9e88bde2fae19b6fb25/0clr54wj5Ly-RkYU9OXPp.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.09802.md","query":{}}">

Papers

arxiv:2608.09802

SWE-Bench ProMax: Benchmarking Agents on Large-Scale Multilingual Code Refactoring

Published on Aug 10

· Submitted by

Yuling on Aug 11

Authors:

Abstract

SWE-Bench ProMax is a rigorously curated multilingual benchmark of large-scale code refactoring tasks that reveals substantial unsolved challenges for current AI coding agents.

Generated by thinkingmachines/Inkling-Small

As AI coding agents take on increasingly complex, long-horizon software engineering tasks, existing benchmarks are rapidly saturating and their evaluation quality has come under serious scrutiny: a recent audit found that nearly 60% of unsolved SWE-bench Verified instances contain flawed tests -- either overly narrow tests that reject correct solutions or overly broad tests that check unstated requirements -- and that frontier models can verbatim reproduce gold patches from training data. Code refactoring, which requires coordinated, behavior-preserving changes across many files, offers a substantially harder and more realistic test of agent capability, yet remains underserved by current benchmarks. We introduce SWE-Bench ProMax, an expert-curated, multilingual code refactoring benchmark of 170 instances drawn from real commits across seven programming languages (Python, Java, TypeScript, Go, C, C++, and Rust). Every instance undergoes rigorous, multi-stage curation that directly addresses the quality problems identified in prior benchmarks: issue descriptions are rewritten from scratch to provide precise, unambiguous specifications, and test suites are manually reviewed to remove overly narrow and overly broad tests. Tasks with insufficient complexity or limited cross-file scope are filtered out, yielding a benchmark of challenging, large-scale refactoring tasks that average 11.4 modified files and 261.6 lines of code per instance, substantially exceeding the scale of existing benchmarks. Experiments with frontier models under two agent scaffolds show that the best model achieves only 41.2% resolve rate, confirming that SWE-Bench ProMax presents a meaningful and unsaturated challenge for current AI coding agents. Our benchmark is available at https://huggingface.co/datasets/swe-bench-promax/SWE-Bench-ProMax.

View arXiv page View PDF Project page Add to collection