{
 "generated": "2026-11-06",
 "sha": "78bbd39a8828",
 "node": {
  "id": "event:sig-0093",
  "type": "Event",
  "name": "METR 高权限实测：Anthropic 的模型能力提升速率已是历史的约 2 倍，大半要归因于用 AI 做研发",
  "slug": "sig-0093",
  "kind": "signal",
  "kind_zh": "行业信号",
  "date": "2026-11-06",
  "summary": "11-05 发布。METR 于 2026-08-12 至 10-29 进驻 Anthropic 做高权限评估，拿到预训练/RL/各领域训练算力、内部模型跑分、AI 加速分析与员工访谈。结论：截至 2026 年 10 月，Anthropic 在明确定义任务上的能力提升速率约为 2024-03 至 2025-10 的两倍（以 AECI 即 Anthropic 版 Epoch 能力指数衡量为 1.9 倍，95% 置信区间 1.5-2.8 倍），相当于一年走完历史两年的进度；并确认这不是训练算力扩张加速带来的，大部分要归因于 Anthropic 大量用 AI 做研发；没有看到明显瓶颈或强递减，因此后续能力进步有相当风险进一步自我加速。METR 同时公开抗议：Anthropic 对其结论最关键的证据做了大面积删节（理由是避免加速其他开发者），公众只能依赖 METR 的判断而无法自行核验。",
  "url": "https://metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment/",
  "evidence": [
   {
    "kind": "own",
    "label": "本站页面",
    "url": "https://xiaopingfeng.com/graph/signals/#sig-0093",
    "text": "AI 行业信号流（Benchmark）：METR 高权限实测：Anthropic 的模型能力提升速率已是历史的约 2 倍，大半要归因于用 AI 做研发"
   },
   {
    "kind": "other",
    "url": "https://metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment/",
    "text": "信号的原文出处（METR）"
   }
  ],
  "entry": "derived",
  "category": "benchmark",
  "category_zh": "Benchmark",
  "source_name": "METR",
  "source_id": "metr",
  "links": [
   {
    "label": "METR",
    "url": "https://metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment/"
   }
  ],
  "degree": 0
 },
 "page": "https://xiaopingfeng.com/graph/e/event/sig-0093/",
 "out": [
  {
   "type": "placed_in",
   "sentence": "METR 高权限实测：Anthropic 的模型能力提升速率已是历史的约 2 倍，大半要归因于用 AI 做研发出现在 Landscape 的「Benchmark」",
   "direction": "out",
   "id": "layer:ai-industry-signals--1",
   "name": "Benchmark",
   "status": "derived",
   "evidence": [
    {
     "kind": "own",
     "label": "本站页面",
     "url": "https://xiaopingfeng.com/graph/signals/#sig-0093",
     "text": "AI 行业信号流（Benchmark）：METR 高权限实测：Anthropic 的模型能力提升速率已是历史的约 2 倍，大半要归因于用 AI 做研发"
    }
   ]
  },
  {
   "type": "involves",
   "sentence": "METR 高权限实测：Anthropic 的模型能力提升速率已是历史的约 2 倍，大半要归因于用 AI 做研发涉及METR",
   "direction": "out",
   "id": "org:metr",
   "name": "METR",
   "role": "issuer",
   "note": "信源：METR",
   "status": "derived",
   "evidence": [
    {
     "kind": "own",
     "label": "本站页面",
     "url": "https://xiaopingfeng.com/graph/signals/#sig-0093",
     "text": "AI 行业信号流（Benchmark）：METR 高权限实测：Anthropic 的模型能力提升速率已是历史的约 2 倍，大半要归因于用 AI 做研发"
    }
   ]
  },
  {
   "type": "mentions",
   "sentence": "METR 高权限实测：Anthropic 的模型能力提升速率已是历史的约 2 倍，大半要归因于用 AI 做研发提到Anthropic",
   "direction": "out",
   "id": "org:anthropic",
   "name": "Anthropic",
   "status": "derived",
   "evidence": [
    {
     "kind": "own",
     "label": "本站页面",
     "text": "名称匹配：「Anthropic」出现在信号的标题里",
     "url": "https://xiaopingfeng.com/graph/signals/#sig-0093"
    }
   ]
  }
 ],
 "in": []
}