Unverified50% confidenceOpinionExact time
The total of 2167 multiple-choice questions and 999 experimental records is still insufficient for training a large-scale generative model, and is better suited as an evaluation benchmark or fine-tuning seed.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/3/2026
First Seen
Valid until: 2/1/2027
Sources
Related Entities
Related Claims
UnverifiedWhen sampling eight candidates without constraints from the model, 10.2% of inputs cannot produce any valid output under limited training data, and this proportion drops to 0.4% when using the full training data77% similarUnverified2022年Hoffman等人的Chinchilla论文重新校准了算力最优的参数-数据比例，指出此前许多大模型训练不充分75% similarUnverifiedThe 29-million-parameter encoder trained by the researchers achieved a popularity-weighted cross-fold AUC of only 0.486 on 985 statements, indistinguishable from the corpus prior.73% similarUnverifiedThe quality ceiling of weak supervision labels depends on the accuracy of the zero-shot model itself; if the model makes systematic errors, the training set will inherit that bias.72% similarUnverified该测试样本规模相对有限（5个模型、13个代码库），结论有待更大规模验证72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1129078API
curl https://kongchang.com/api/v1/knowledge/claims/1129078MCP
get_claim(id=1129078)