Unverified50% confidenceFactExact time
训练使用约500万行英文维基百科语料约3亿token，10万上下文词表，23758个名词目标
1
Sources
50%
Confidence
Long-term
Relevance
7/13/2026
First Seen
Sources
No MLP or Attention Required: How Point-Attractor Dynamics Learn Word Embeddings
redditr/learnmachinelearning7/10/2026
Related Claims
UnverifiedWhisper's training data contains 680,000 hours, largely sourced from videos on the internet with varying subtitle quality.67% similarUnverified深度研究单次任务消耗的token数量可达普通对话的数十倍，成本可能相当于数十次乃至上百次普通问答的资源消耗65% similarUnverifiedCommon Crawl是非营利组织运营的大规模网络爬取项目，每月抓取数十亿网页，数据集被用于训练包括GPT系列在内的大型语言模型64% similarUnverifiedThe build process involved 15,314 model calls, more than 339 million input tokens, and a total of more than 2.6 billion tokens (including cached reads, output, and thinking tokens).64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/496331API
curl https://kongchang.com/api/v1/knowledge/claims/496331MCP
get_claim(id=496331)