Unverified50% confidenceFactExact time

GPT-3 的训练数据中约60%来自经过过滤的Common Crawl

1

Sources

50%

Confidence

Long-term

Relevance

8/16/2026

First Seen

Sources

Terminal Bench 3发布：未被污染的AI终端能力基准测试

Related Claims

UnverifiedGPT-5.5为了更易消化，在训练中牺牲了对细节的洞察力，评测者建议需要敏锐洞察力的工作仍使用Opus 4.774% similar UnverifiedGPT-5.5的训练数据量达到了约45PB72% similar Unverified学术界多项研究表明GPT-4、LLaMA等主流模型在MMLU、HumanEval、GSM8K等榜单上的高分有相当比例来自训练集与测试集内容重叠72% similar Unverified谷歌DeepMind的研究发现GPT类模型在特定提示条件下可精确复现训练集中约1%的文本片段71% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/757584

API

curl https://kongchang.com/api/v1/knowledge/claims/757584

MCP

get_claim(id=757584)