Unverified50% confidenceFactExact time
Under a second-order approximation, LIRA estimates the magnitude of local loss perturbation caused by recent LoRA factor movements, using this as the score for each layer.
1
Sources
50%
Confidence
Long-term
Relevance
11/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedLoss correlation refers to the Pearson correlation coefficient between the two loss curves during training; 0.99 means that at each training step the NPU-side loss value rises and falls almost in sync with the GPU side.67% similarUnverifiedGGA allocates bit widths to different heads under a model-level mixed-precision budget by evaluating the gradient impact of each attention head's candidate quantization residual on the future prediction loss.64% similarUnverifiedThe LoRA baseline used for comparison in the paper uses rank 19 with 6.6 million parameters, achieving a correction rate of 83.3%, but exhibited capability loss of 30% to 75%.63% similarUnverifiedThe loss during training is calculated per token, and long samples contribute far more gradient than short samples, so a 10% share of rows can absolutely correspond to more than 50% of the token volume and gradient signal.61% similarUnverifiedAs model scale increases, the accuracy loss from using DA shows a trend of diminishing.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1133707API
curl https://kongchang.com/api/v1/knowledge/claims/1133707MCP
get_claim(id=1133707)