Diagnostic analysis of quantization errors in speaker verification models reveals a 2-bit knee point and supports a calibrated cascade that maintains FP32-level scores at reduced compute cost.
Towards lightweight speaker verification via adaptive neural network quantization,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
On Low-Bit Quantization Errors in Speaker Verification: Diagnostic and Mitigation
Diagnostic analysis of quantization errors in speaker verification models reveals a 2-bit knee point and supports a calibrated cascade that maintains FP32-level scores at reduced compute cost.