ReQuant refines already-quantized language models on a fixed integer grid via backpropagation-free coordinate descent, consistently lowering perplexity, KL divergence, and zero-shot accuracy gaps across models, bit-widths, and PTQ initializers.
Gomez, Łukasz Kaiser, and Illia Polosukhin
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
ReQuant: Fixed-Grid Discrete Refinement for Post-Training Quantization
ReQuant refines already-quantized language models on a fixed integer grid via backpropagation-free coordinate descent, consistently lowering perplexity, KL divergence, and zero-shot accuracy gaps across models, bit-widths, and PTQ initializers.