Unverified50% confidenceFactExact time
The momentum method introduces a velocity variable v that accumulates historical gradients: v = βv + (1-β)∇θL, with the parameter update θ = θ - αv, where β is typically set to 0.9.
1
Sources
50%
Confidence
Long-term
Relevance
11/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn Strata's speed table, under a 128k context, Q2_0 (2-bit) achieves 65.1 tokens/s and IQ2_XS achieves 52 tokens/s; under a 4k short context, IQ3_XXS achieves 65.6 tokens/s.63% similarUnverifiedFor Gaussian arms with a common variance, H₁-SR achieves at least 11/12 of the optimal normalized rate among all adaptive algorithms, and this ratio holds for any K≥2.59% similarUnverifiedWhen scanning thresholds in steps of 0.01, precision rises while recall falls, the two cross near a threshold of 0.63 (P≈0.733, R≈0.725), and the F1 peak is at a threshold of 0.41 (F1≈0.758).59% similarUnverifiedIf the optimizer itself is 1.5 times faster, in the case where the optimizer accounts for 10%, the end-to-end speedup is about 1/(0.9+0.1/1.5)≈1.03, that is, only about 3%.57% similarUnverifiedOn the 1,302 expert-annotated trajectories of AgentRewardBench, JEV-27B-VL reports an AUROC of 0.91, with precision 78.4 and recall 70.2 at a threshold of 0.5.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1126400API
curl https://kongchang.com/api/v1/knowledge/claims/1126400MCP
get_claim(id=1126400)