Aug 2026
Q-SA2: Enhanced low-bit quantization-aware fine-tuning via structure-aware adaptation.
The proposed Structure-Aware Adaptation for low-bit QAF (Q-SA2) extends to a weight-wise mixed-precision variant with a distribution-based outlier detection strategy on sparse components, enabling Q-SA2 to adaptively allocate higher precision to critical weights.
Yuning Yang, Guo-Wei Peng, Xiurui Xie et al.
· Neural Networks · 0 citations