Unverified50% confidenceFactExact time
Distribution shift can cause significant drops in model performance or even complete failure
1
Sources
50%
Confidence
Long-term
Relevance
10/22/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWhen the success rate is extremely low, almost all sample rewards are identical, resulting in extremely high policy gradient variance and very low signal-to-noise ratio, preventing the model from obtaining effective update directions.71% similarUnverified分层模型路由策略若分类不准确或过度偏向成本控制，会导致用户感知到的质量下降70% similarUnverifiedFor cell types or species underrepresented in the training data, the reliability of predictions from such models decreases significantly.70% similarUnverifiedCompared to full fine-tuning, which requires updating all parameters of the model, LoRA significantly reduces GPU memory usage.70% similarUnverifiedAs model scale increases, the accuracy loss from using DA shows a trend of diminishing.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1083779API
curl https://kongchang.com/api/v1/knowledge/claims/1083779MCP
get_claim(id=1083779)