이번 글에서는 Specification-first convergence with an AI coding agent: a case study of dismantling a core architectural invariant across 189 files in a 717k-line codebase with no test oracle and no human code review 논문의 핵심 포인트만 간단히 정리한다. 2026년 8월 12일(Arxiv) Abenhaim, Joel. AI Sovereign Labs, Paris, France 논문...

This article explains the key points of Agent Safety Should Be a Runtime Contract. 2026-08-11 (arXiv), Preprint Ng, Albus W., Han, Yi, Zhang, Jusheng, Wang, Wenhao. Vast Intelligence Lab, Southwest University, Sun Yat-sen University Paper Read this article in Korean Summary This position paper argues that safety for consequential AI...

이번 글에서는 Agent Safety Should Be a Runtime Contract 논문의 핵심 포인트만 간단히 정리한다. 2026년 8월 11일(Arxiv), Preprint Ng, Albus W., Han, Yi, Zhang, Jusheng, Wang, Wenhao. Vast Intelligence Lab, Southwest University, Sun Yat-sen University 논문 링크 영문판 보기 요약 이 입장 논문은 중요한 결과를 초래하는 AI agent의 안전성을 학습...

This article explains the key points of SWE-Bench ProMax: Benchmarking Agents on Large-Scale Multilingual Code Refactoring. 2026-08-10 (arXiv), COLM 2026 Shi, Yuling, Xu, Jinghan, Fu, Kelin, Zeng, Wenhao, He, Shilin, Zhang, Lei, Liu, Yue, Zhao, Zelin, Zhuo, Terry Yue, Cao, Jialun, et al. Shanghai Jiao Tong University, Peking University, The...

이번 글에서는 SWE-Bench ProMax: Benchmarking Agents on Large-Scale Multilingual Code Refactoring 논문의 핵심 포인트만 간단히 정리한다. 2026년 8월 10일(Arxiv), COLM 2026 Shi, Yuling, Xu, Jinghan, Fu, Kelin, Zeng, Wenhao, He, Shilin, Zhang, Lei, Liu, Yue, Zhao, Zelin, Zhuo, Terry Yue, Cao, Jialun, et al. Shanghai Jiao Tong University, Peking University,...