Unverified50% confidenceTradeoffExact time
Overly aggressive biasing can often reduce B-WER but simultaneously raises U-WER, resulting in limited or even negative improvement in overall WER.
1
Sources
50%
Confidence
Long-term
Relevance
9/25/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWhile PTC-Bias reduces B-WER, U-WER (unbiased word error rate) remains largely unchanged.69% similarUnverifiedThe tendency to over-scope may be an advantage in complex backend tasks, but lowers the overall score in scenarios requiring precise, restrained modifications.61% similarUnverifiedGradient descent naturally tends to select flatter, simpler solutions from among the many solutions that perfectly fit the training set, and this implicit regularization effect is closely related to generalization ability.60% similarUnverifiedPlasticity decay refers to the phenomenon where, as a model undergoes more and more fine-tuning updates, its weights gradually solidify and its ability to respond to new information continuously declines.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/949151API
curl https://kongchang.com/api/v1/knowledge/claims/949151MCP
get_claim(id=949151)