Unverified50% confidenceFactExact time
In real models trained on the same data, the correlation between BPB and downstream task performance is weak.
1
Sources
50%
Confidence
Long-term
Relevance
11/4/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedDuring the post-training phase, an excessively low loss may correspond to overfitting, where the model memorizes the training data and loses generalization ability.68% similarUnverifiedThe Rif and Kabyle variety subsets are extremely small in scale, which is a significant shortcoming for training models that cover all varieties.66% similarUnverifiedIn real-world scenarios, the source training set and the target test set are rarely identically distributed.66% similarUnverifiedThe score improvement in the first round of random task training was very small, because a large amount of training time was wasted on samples the model stably got right or wrong.66% similarUnverifiedThe paper finds that the ranking of downstream tasks is not consistent with the ranking of training loss, and schemes with lower training loss do not necessarily perform better in downstream evaluations.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1134078API
curl https://kongchang.com/api/v1/knowledge/claims/1134078MCP
get_claim(id=1134078)