AntSpeaker/MECT-B2

Source and architecture

• Displayed name: AntSpeaker/MECT-B2
• Submitted by: Public checkpoint
• Metadata: OpenSVBench model record
• Source: AntSpeaker
• Architecture: MECT-B2
• Model family: MECT
• Submitted alias: AntSpeaker/MECT-B2
• Created at: 2026-10-19T05:05:41.317553+00:00

Training data and references

• Training data: VoxCeleb2 dev + VoxBlink2
• Training setup: MECT-B2 uses 80-dimensional mean-normalized log-Mel filterbanks at 16 kHz with a 25 ms window, 10 ms shift, and 20–7600 Hz band, followed by a four-stage CNN-Transformer with C=32, M=3, and Transformer width d=64. The model uses frame-level dense mixture-of-experts with four experts, weighted dense fusion, attentive statistics pooling, and a 192-dimensional projection trained with SphereFace2. The VoxCeleb2 + VoxBlink2 setting uses full training followed by large-margin fine-tuning without speed perturbation; the LMF stage uses VoxCeleb2.
• Submission mode: full-core
• Model source: Open link
• Paper: MECT: Mixture of Experts with CNN-Transformer Network for Speaker Verification

Scenario ranks above the global position

Scenario ranks below the global position