ConceptGrouped Sparse Quantization / Residual Compression and Optimization
RCO-GSQ
A low-bit quantization scheme for large language models proposed by ISTA-DASLab, combining grouped sparse quantization with residual compression optimization. At extremely low precisions such as 2-bit, it employs an error compensation mechanism to maintain acceptable model quality.
Timeline (last 90 days)
Sep 24
RCO-GSQ is a low-bit quantization scheme proposed by ISTA-DASLab, where GSQ introduces sparsity through weight grouping to reduce precision loss, and RCO performs further compression and correction on the quantization residuals.
Unverified50%