AWhat changed
Co-designed affine scale/zero-point quantization, 8-bit weight/activation and 32-bit accumulation/bias arithmetic with floating-point fake-quantization training for efficient mobile inference.
Scientific scope
Co-designed affine scale/zero-point quantization, 8-bit weight/activation and 32-bit accumulation/bias arithmetic with floating-point fake-quantization training for efficient mobile inference. Integer-only refers to deployed inference, not all training or offline scale preparation. Inherits earlier low-precision/fixed-point inference; rounding, zero-point correction, fused-operator placement and hardware kernels matter. Evaluated mobile CNN latency/accuracy tradeoffs do not guarantee lossless quantization or speedups on arbitrary models/devices.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Integer-only refers to deployed inference, not all training or offline scale preparation. Inherits earlier low-precision/fixed-point inference; rounding, zero-point correction, fused-operator placement and hardware kernels matter. Evaluated mobile CNN latency/accuracy tradeoffs do not guarantee lossless quantization or speedups on arbitrary models/devices.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction