Unverified50% confidenceFactExact time
The fine-tuning used 4,260 training samples and 474 evaluation samples, all of which were STEM question-and-answer data with chain-of-thought.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe research team used the scorer to score 322,250,000 FineWeb-Edu-Fortified documents and constructed a curated pre-training data mixture.72% similarUnverifiedThe research team conducted controlled experiments on four trained DeBERTa checkpoints across a total of 7,890 claims.71% similarUnverifiedThe 503-question analysis set itself includes training questions, so the results are partly transductive, which gives the planner a certain advantage.68% similarUnverifiedEleutherAI的Pythia项目同时公开了训练数据、代码和完整检查点序列，包含从70M到12B参数的8个规模梯度模型，全部在相同的Pile数据集上以相同顺序训练，并公开了超过150个检查点68% similarUnverifiedThe research team selected 8 models spanning different capability tiers and conducted cross-evaluations across 54 benchmarks67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/999486API
curl https://kongchang.com/api/v1/knowledge/claims/999486MCP
get_claim(id=999486)