MARR uses per-module adaptive residual scaling updated by PID feedback to balance error correction against Hessian-approximation bias in low-bit PTQ.
Owq: Outlier-aware weight quantization for efficient fine-tuning and inference of large language models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
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.