{
 "generated": "2026-11-05",
 "sha": "707df050bf85",
 "node": {
  "id": "ep:30",
  "type": "Episode",
  "name": "EP.30",
  "slug": "30",
  "ep": 30,
  "date": "2025-02-28",
  "page": "/buzzwords/30/",
  "n_models": 0,
  "title": "别再罗列 MMLU，直接说测的是什么能力",
  "oneliner": "把 benchmark 名字换成人话是可以直接抄的产品设计；SmolVLM 2 真正跑进端侧；Claude 3.7 在 coding 上很强",
  "minutes": 56,
  "template": "archive",
  "status": "published",
  "kind": "archive",
  "kind_zh": "档案重建",
  "entry": "agent",
  "by": "episode-import",
  "at": "2026-11-04T16:15:12",
  "n_stories": 5,
  "n_topics": 5,
  "n_items": 8,
  "n_sections": 1,
  "n_material": 8,
  "degree": 0,
  "n_entities": 8,
  "scale": "5 个故事线 / 专题 · 8 个条目 · 讲到 8 个实体"
 },
 "page": "https://xiaopingfeng.com/graph/e/ep/30/",
 "out": [],
 "in": [
  {
   "type": "part_of",
   "sentence": "Claude 3.7 在 coding 上确实很强属于EP.30",
   "direction": "in",
   "id": "story:30-t1",
   "name": "Claude 3.7 在 coding 上确实很强",
   "status": "derived",
   "evidence": []
  },
  {
   "type": "part_of",
   "sentence": "SmolVLM 2：视觉模型真的跑进端侧了属于EP.30",
   "direction": "in",
   "id": "story:30-t2",
   "name": "SmolVLM 2：视觉模型真的跑进端侧了",
   "status": "derived",
   "evidence": []
  },
  {
   "type": "part_of",
   "sentence": "NotebookLM 的前创始成员出来创业了属于EP.30",
   "direction": "in",
   "id": "story:30-t3",
   "name": "NotebookLM 的前创始成员出来创业了",
   "status": "derived",
   "evidence": []
  },
  {
   "type": "part_of",
   "sentence": "开源面试官：值得看的是它的 system prompt属于EP.30",
   "direction": "in",
   "id": "story:30-t4",
   "name": "开源面试官：值得看的是它的 system prompt",
   "status": "derived",
   "evidence": []
  },
  {
   "type": "part_of",
   "sentence": "把基准名字换成人话属于EP.30",
   "direction": "in",
   "id": "story:30-t5",
   "name": "把基准名字换成人话",
   "status": "derived",
   "evidence": []
  },
  {
   "type": "part_of",
   "sentence": "@FinanceYF5：最近AI新闻：Anthropic刚刚发布了世界上首个“混合推理”AI模型。 此外，还有来自阿里巴巴Qwen、Gibber Link、人形开发公司1X Technologies、谷歌、OpenAI、DeepSeek…属于EP.30",
   "direction": "in",
   "id": "item:30-m1-4c8ffa",
   "name": "@FinanceYF5：最近AI新闻：Anthropic刚刚发布了世界上首个“混合推理”AI模型。 此外，还有来自阿里巴巴Qwen、Gibber Link、人形开发公司1X Technologies、谷歌、OpenAI、DeepSeek…",
   "status": "derived",
   "evidence": []
  },
  {
   "type": "part_of",
   "sentence": "A Practical Guide to Implementing DeepSearch/DeepResearch属于EP.30",
   "direction": "in",
   "id": "item:30-m4-c3b9d2",
   "name": "A Practical Guide to Implementing DeepSearch/DeepResearch",
   "status": "derived",
   "evidence": []
  },
  {
   "type": "part_of",
   "sentence": "@akshay_pachaar：Before we begin, here’s what we’ll be doing: We’ll fine-tune our private and locally running DeepSeek-…属于EP.30",
   "direction": "in",
   "id": "item:30-m4-d9be81",
   "name": "@akshay_pachaar：Before we begin, here’s what we’ll be doing: We’ll fine-tune our private and locally running DeepSeek-…",
   "status": "derived",
   "evidence": []
  },
  {
   "type": "part_of",
   "sentence": "官宣！OpenAI前CTO新公司：北大校友翁荔加盟，创始29人2/3来自OpenAI属于EP.30",
   "direction": "in",
   "id": "item:30-m5-c4b15b",
   "name": "官宣！OpenAI前CTO新公司：北大校友翁荔加盟，创始29人2/3来自OpenAI",
   "status": "derived",
   "evidence": []
  },
  {
   "type": "part_of",
   "sentence": "Automating GPU Kernel Generation with DeepSeek-R1 and Inference Time Scaling属于EP.30",
   "direction": "in",
   "id": "item:30-m7-68fde4",
   "name": "Automating GPU Kernel Generation with DeepSeek-R1 and Inference Time Scaling",
   "status": "derived",
   "evidence": []
  },
  {
   "type": "part_of",
   "sentence": "OpenAI推出真实世界百万报酬AI编程能力测试基准：实测Claude 3.5 最强！属于EP.30",
   "direction": "in",
   "id": "item:30-m8-af90ca",
   "name": "OpenAI推出真实世界百万报酬AI编程能力测试基准：实测Claude 3.5 最强！",
   "status": "derived",
   "evidence": []
  },
  {
   "type": "part_of",
   "sentence": "GitHub - openai/SWELancer-Benchmark: This repo contains the dataset and code for the paper \"SWE-Lancer: Can Frontier L…属于EP.30",
   "direction": "in",
   "id": "item:30-m8-2c47dd",
   "name": "GitHub - openai/SWELancer-Benchmark: This repo contains the dataset and code for the paper \"SWE-Lancer: Can Frontier L…",
   "status": "derived",
   "evidence": []
  },
  {
   "type": "part_of",
   "sentence": "这里领取属于EP.30",
   "direction": "in",
   "id": "item:30-m9-198455",
   "name": "这里领取",
   "status": "derived",
   "evidence": []
  }
 ]
}