[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f1g52d5ww321w":3},{"slug":4,"category":5,"publishedAt":6,"titleZh":7,"titleEn":8,"summaryZh":9,"summaryEn":10,"models":11,"hardwares":15,"bodyZh":19,"bodyEn":20,"sourceName":21,"sourceUrl":22,"updatedAt":23},"maya-s-deepseek41-thinking-contaminated-report-2026-11-13","ARENA","2026-11-12T23:12:30.643Z","DeepSeek Maya-S 社区追测：思考档分数更高，但测试自己标了污染","DeepSeek Maya-S follow-up: higher thinking score comes with a contamination flag","11月12日一份双3090 Ti报告补充150项能力测试，思考关闭\u002F开启的pass@1为119\u002F131；开启档被工具标为CONTAMINATED，不能当干净思考收益或模型排名。本站仅核对原文，未独立复现。","A November 12 dual-3090-Ti report adds a 150-scenario test: pass@1 is 119 with thinking off and 131 with it on. The harness flags the thinking arm CONTAMINATED. It is not a clean thinking benefit or a model ranking; AraoAI has not reproduced it.",[12],{"id":13,"name":14},85,"DeepSeek-V4.1-Flash-1017",[16],{"id":17,"name":18},115,"2x NVIDIA RTX 3090 Ti","11月12日，club-3090 issue #2009报告了DeepSeek-V4.1-Flash-1017 Maya-S在双RTX 3090 Ti上的运行，并在20:32 UTC补交能力测试。最需要读的不是更高的总分，而是作者同时公开的有效性标记：思考开启档被测试工具标为 CONTAMINATED。\n\n1. 把机器和配置说清。作者使用Project Maya v1.1.57、TP=2、NVLink及DSpark草稿模型，GPU功耗上限每张325W，宿主环境是KVM虚拟机并直通GPU。它不是双3090的默认功耗成绩，也不是只有显存的纯GPU运行；报告明确有大量专家放在主机内存中。不能把这份结果拿去证明任意双卡都一样。\n\n2. 总分提升不等于干净因果。补充测试每档只跑一轮、覆盖150个场景，pass@1为关闭119、开启131；pass@3则为134与136。开启档在toolcall-15中仅5个响应返回reasoning，因此工具将该档标为污染，而不是完整受控的思考对照。不能把12项差值说成稳定收益，也不能把重试成绩当首答成绩。\n\n3. 测试条件不能省略。采样继承服务器temperature 1.0、top_p 1.0，不是工具默认的温度零；作者明说不可与温度零基线横比。思考开启档还有两项token_limit，沙箱镜像落后于最近的测试工具更新。后续补测了能力，不代表verify-full、verify-stress与持续soak也做完；评论明确这些仍未运行。\n\n同帖也保留了短重复提示与新文本速度不同、不同引擎版本及互联条件不可直接比较的说明。采购或部署选型应索要原始日志、准确权重、引擎构建、采样协议与失败清单，再判断证据能回答哪个问题。单份自报能提供复现线索，不能自动升级成跨机器验证。本文不发布模型排名，也没有本站独立能力成绩。\n\n原报告：https:\u002F\u002Fgithub.com\u002Fnoonghunna\u002Fclub-3090\u002Fissues\u002F2009\n追测：https:\u002F\u002Fgithub.com\u002Fnoonghunna\u002Fclub-3090\u002Fissues\u002F2009#issuecomment-6835193065","On November 12, club-3090 issue #2009 reported DeepSeek-V4.1-Flash-1017 Maya-S on two RTX 3090 Ti cards, followed by a quality-test comment at 20:32 UTC. The important detail is not simply a higher total: the harness itself marked the thinking-on arm CONTAMINATED.\n\n1. Preserve the configuration. The author used Project Maya v1.1.57, tensor parallelism of two, NVLink and the DSpark drafter, with each GPU capped at 325 W. The environment was a KVM guest with GPU passthrough and substantial host-RAM expert offload. These are not stock-power dual-3090 measurements or a GPU-only deployment, and they do not establish a result for arbitrary dual-card systems.\n\n2. A higher total is not a clean causal result. Each arm ran once over 150 scenarios. Pass@1 was 119 with thinking off and 131 with it on, while pass@3 was 134 and 136. Only five of fifteen tool-calling responses in the thinking-on arm contained reasoning, triggering the contamination flag. The twelve-point difference is not a demonstrated stable thinking benefit, and retry-aware scores are not first-answer scores.\n\n3. The protocol matters. Sampling inherited server temperature 1.0 and top_p 1.0 rather than the harness temperature-zero default; the author explicitly warns against comparison with that baseline. Two thinking-on cases hit token limits, and sandbox images predated the latest harness update. The follow-up adds quality testing, not completed verify-full, verify-stress or continuous soak, which the comment still lists as unrun.\n\nThe report also distinguishes repeated short prompts from fresh-text speed and warns about comparing different engines, power caps and interconnects. For selection decisions, request raw logs, exact weights, engine identity, sampling protocol and failures. A single self-report is a reproduction lead, not cross-machine validation. AraoAI has checked the source but has not independently measured these results.\n\nReport: https:\u002F\u002Fgithub.com\u002Fnoonghunna\u002Fclub-3090\u002Fissues\u002F2009\nFollow-up: https:\u002F\u002Fgithub.com\u002Fnoonghunna\u002Fclub-3090\u002Fissues\u002F2009#issuecomment-6835193065","club-3090 community report \u002F follow-up","https:\u002F\u002Fgithub.com\u002Fnoonghunna\u002Fclub-3090\u002Fissues\u002F2009","2026-11-12T23:12:30.644Z"]