Unverified50% confidenceFactExact time
Macro-F1 assigns equal weight to each category by computing the F1 score for each category separately and then taking a simple average, making it suitable for scenarios with imbalanced class distributions.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedHakemBench evaluates models across three dimensions—decision quality (macro F1), calibration (normalized Brier score), and selective automation (normalized area under the generalized risk-coverage curve)—and combines them into a composite score via geometric mean.57% similarUnverifiedThe paper proposes a covariate-dependent continuous ratio Markov random field model that directly performs joint modeling of multiple ordinal attributes and introduces covariates so that the preference distribution varies with user characteristics or context.57% similarUnverifiedStatic quantization makes fixed quantization level choices for certain types of tensors, while dynamic quantization makes differentiated choices for each tensor at each checkpoint size.56% similarUnverifiedEvaluation metrics use each model's own normal prose output as the baseline, and each model's own tokenizer is used to calculate token counts to ensure fairness of comparison.56% similarUnverifiedIf the weight matrices after training exhibit a heavy-tailed spectral distribution, this is often a strong indicator of better model quality.56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1001318API
curl https://kongchang.com/api/v1/knowledge/claims/1001318MCP
get_claim(id=1001318)