A quantization vector derived from a donor model via weight-space arithmetic can be added to a receiver model to improve post-PTQ Top-1 accuracy by up to 60 points in 3-bit settings without receiver-side QAT or data.
Low-bit model quantization for deep neural networks: A survey.arXiv preprint arXiv:2505.05530
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 4verdicts
UNVERDICTED 4roles
background 1polarities
background 1representative citing papers
DECO is a sparse MoE architecture with ReLU-based routing, learnable expert scaling, and NormSiLU activation that matches dense Transformer performance at 20% expert activation and delivers 2.93x speedup on Jetson AGX Orin.
MARR uses per-module adaptive residual scaling updated by PID feedback to balance error correction against Hessian-approximation bias in low-bit PTQ.
Tiny NeRV models using capacity scaling, frequency-aware distillation, and low-precision quantization achieve favorable quality-efficiency trade-offs with far fewer parameters and lower computational costs than standard NeRV.
citing papers explorer
-
Zero-Shot Quantization via Weight-Space Arithmetic
A quantization vector derived from a donor model via weight-space arithmetic can be added to a receiver model to improve post-PTQ Top-1 accuracy by up to 60 points in 3-bit settings without receiver-side QAT or data.
-
DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices
DECO is a sparse MoE architecture with ReLU-based routing, learnable expert scaling, and NormSiLU activation that matches dense Transformer performance at 20% expert activation and delivers 2.93x speedup on Jetson AGX Orin.
-
MARR: Module-Adaptive Residual Reconstruction for Low-Bit Post-Training Quantization
MARR uses per-module adaptive residual scaling updated by PID feedback to balance error correction against Hessian-approximation bias in low-bit PTQ.
-
TinyNeRV: Compact Neural Video Representations via Capacity Scaling, Distillation, and Low-Precision Inference
Tiny NeRV models using capacity scaling, frequency-aware distillation, and low-precision quantization achieve favorable quality-efficiency trade-offs with far fewer parameters and lower computational costs than standard NeRV.