[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"aihot-art-92903":3},{"itemId":4,"vertical":5,"category":6,"source":7,"score":8,"title":9,"summary":10,"analysis":11,"url":12,"coverUrl":13,"direction":13,"marketSignal":13,"publishedAt":14},"92903","ai","论文研究","Rohan Paul",63,"Kimi K3在DeepSWE基准测试中排名第三，与Claude Fable和GPT-5.6 Sol并列","Kimi K3在DeepSWE基准测试中排名第三，与Claude Fable和GPT-5.6 Sol并列，成为首个达此成绩的开源模型。","Kimi K3 在 DeepSWE 基准测试中排名第三，与 Claude Fable 和 GPT-5.6 Sol 并列，成为首个达到此成绩的开源权重模型。\n· 得分从 K2.7-code 的 31 分跃升至 69 分，提升 38 分，进步显著。\n· DeepSWE 测试智能体在真实代码库中完成多文件修改的能力，是衡量端到端软件工程能力的强信号。\n· 开源模型首次与顶尖闭源模型并列，意义重大。\n影响\u002F看点：开源模型在软件工程任务上取得突破，缩小了与闭源模型的差距。","https:\u002F\u002Fx.com\u002Frohanpaul_ai\u002Fstatus\u002F2078598860801520109",null,"2026-07-19 05:52:41"]