Unverified50% confidenceBenchmarkExact time
Marimba's binary vocabulary judgment has a maximum ECE of 0.08.
1
Sources
50%
Confidence
Long-term
Relevance
11/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMarimba achieves 0.93 accuracy on the four-class Speech Commands task.64% similarUnverifiedOn calibration, Jev's ECE in independent tests varies between 0.02 and 0.35 depending on the task; Laya's ECE is 0.466 before temperature fitting (overconfident) and drops to 0.081 after fitting; Drex has not published calibration data.63% similarUnverifiedMarimba's turn-end detection readout reaches AUROC 0.930, while the dedicated model Smart Turn v3.2 achieves 0.982 on its own test set.59% similarUnverifiedThe confidence Earlybird gives for the word currently being spoken is calibrated, with expected calibration error (ECE) between 3.7 and 5.5.58% similarUnverifiedMarimba's speaker counting (0/1/2/3+) accuracy exceeds that of pyannote 3.1.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1173844API
curl https://kongchang.com/api/v1/knowledge/claims/1173844MCP
get_claim(id=1173844)