Unverified50% confidenceFactExact time
The paper conducts experiments on the FineWeb-Edu dataset at four model scales, from 60M to 1B parameters.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/21/2026
First Seen
Valid until: 1/19/2027
Sources
Related Entities
Related Claims
VerifiedIn 2022, Hoffmann et al.'s Chinchilla paper redid experiments on four hundred models and concluded that, under the premise of compute-optimality, each parameter should correspond to approximately 20 tokens.72% similarUnverifiedThe researchers collected 4,249 run trajectories from four frontier models, plus 4,486 additional trials from 26 models.71% similarUnverifiedChinchilla论文通过对超过400个不同规模模型的系统性实验得出结论70% similarUnverifiedThe practical tests conducted by this Reddit community covered a range from 1.5B small models to 120B large models.67% similarUnverifiedPew's ATP design effect typically ranges between 1.5 and 2.5, meaning that a survey with a nominal sample size of 1,000 may have an effective sample size of only 400–667.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1078923API
curl https://kongchang.com/api/v1/knowledge/claims/1078923MCP
get_claim(id=1078923)