Sequency-ordering Hadamard rows and grouping them into diagonal blocks reduces 2-bit quantization error, giving training-free perplexity of 11.59 versus 10.97 for learned OSTQuant on Llama-2-7B.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Grouped Sequency-arranged Rotation: Optimizing Rotation Transformation for Quantization for Free
Sequency-ordering Hadamard rows and grouping them into diagonal blocks reduces 2-bit quantization error, giving training-free perplexity of 11.59 versus 10.97 for learned OSTQuant on Llama-2-7B.