Jinze Li, Weijie Liu, Jinfeng Xu, Shuo Yang, Xin Yang, Yang Zhang, Yijin Wang, Zifei Shan · 2026 · arXiv
DOI: 10.48550/arXiv.2610.25899 · open-access PDF
소프트웨어 엔지니어링 (SWE) 벤치마크는 코딩 에이전트의 능력을 측정하는 주요 방법입니다 : 에이전트는 코드 저장소 또는 기계 환경에 배치되며, 요청을 받았으며 숨겨진 테스트에 의해 판단됩니다. 이러한 벤치마크는 거의 항상 작업에 대한 이전 작업이없는 깨끗한 상태에서 에이전트를 시작합니다.
Machine translation (Meta M2M100); the English original follows.
Software engineering (SWE) benchmarks are the primary way to measure the capabilities of coding agents: an agent is placed in a code repository or machine environment, given a request, and judged by hidden tests. These benchmarks almost always start the agent from a clean state that contains no earlier work on the task. In practice, however, agents often continue work that a developer or another agent has already done and reported as complete, with no indication of what remains incorrect or missing. We introduce SWE-Residual, a benchmark for Starting-State Robustness: the ability to maintain p
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.25899, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange