Event ·Official first-hand

七款模型端到端做数据科学项目:反馈循环平均分从59%提升到76%

一项研究让 7 个模型(来自 Anthropic、OpenAI、Google、DeepSeek、Z.AI)互为写手和评分者,完成 49 组配对、各跑 3 次的端到端数据科学项目,包括下载数据、清洗、分析和写论文。147 份初始提交中 107 份可运行渲染,平均分从 59% 升至 76%。单个模型的提升幅度从 Anthropic Opus 5.5 的 3 个百分点到 Haiku 4.5 的 46 个百分点不等。模型给自己的作品打分平均比其他模型高约 9 个百分点,且约一半的轮次间提升来自评分者自身的噪声。

Current read

七个模型轮流写、轮流评数据科学项目,谁提升大、谁自评虚高,这篇都量出来了,做评测的朋友值得细看。

11 sources38° AI heatLast updated 2026/11/3 19:39:03

Evidence

11 sources
Primary sourcearXiv: DeepSeek
View source
Related source 1rohanpaul_ai
View source
Related source 2lmarena.ai
View source
Related source 5Tech in Asia
View source
Related source 6shao__meng
View source
Related source 8Bindu Reddy
View source
Related source 9The Information
View source
Related source 10Nathan Lambert
View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.