Unverified50% confidenceFactExact time
2022年DeepMind发表了Chinchilla论文《Training Compute-Optimal Large Language Models》，由Jordan Hoffmann等人撰写，指出在给定计算预算下参数量与训练数据量需保持合理配比
1
Sources
50%
Confidence
Long-term
Relevance
8/28/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedDeepMind's Chinchilla scaling law demonstrates that the optimal ratio of training data volume to model parameter count evolves over time, and equal FLOP investment can produce models with significantly different capabilities74% similarUnverified在代码语料比例更高的混合数据集上训练的模型，在数学推理基准（如MATH、GSM8K）上的表现显著优于纯自然语言训练的同规模模型，这一结论获斯坦福大学和MIT的多项研究证实71% similarUnverifiedThe Scaling Law paper by Kaplan et al. in 2020 revealed that, under a fixed compute budget, there is an optimal ratio between model parameter count and the number of training tokens (the Chinchilla optimal).70% similarUnverified论文在一个10亿参数模型上用2000亿token的Numeratron数学与Dolma混合语料做训练，展示了中训练损失与RL性能之间的关系69% similarUnverifiedThe inference process of large language models consumes as much computational resources as the training process.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/810528API
curl https://kongchang.com/api/v1/knowledge/claims/810528MCP
get_claim(id=810528)