Unverified50% confidenceBenchmarkExact time
Under the multi-pass training mode, overall submission calibration error decreases by 63%–68%.
1
Sources
50%
Confidence
Long-term
Relevance
10/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe pre-training validation loss across the three generations of models shows a steady downward trend: 2.87 → 2.78 → 2.59.75% similarUnverifiedIn CLM training, pretraining alone scores 52% on the hard negative sample test; adding the hard negative sample stage jumps it to 69%; training with hard negative samples from day one peaks at only 62% and overfits quickly; omitting replay causes the hard negative sample score to drop from 69% back to 56%.74% similarUnverifiedAfter removing the candidate filtering component, FEVER accuracy dropped to 76.24%.73% similarUnverified消融实验显示，若继续训练主干通用知识掉10%、数学掉17.8%；若两塔共享权重则通用任务掉26%，其余指标全线崩溃69% similarUnverifiedIn real-environment A/B testing, the post-trained checkpoint reduced the tool-call failure rate by 21.2% relatively compared to the earlier version.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/975562API
curl https://kongchang.com/api/v1/knowledge/claims/975562MCP
get_claim(id=975562)