BOF4-S with outlier preservation computes Lloyd-optimal 4-bit block-wise codebooks and beats NF4 and AF4 on quantization error and perplexity.
QLORA: Efficient Finetuning of Quantized LLMs,
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
-
Improving Block-Wise LLM Quantization by 4-bit Block-Wise Optimal Float (BOF4): Analysis and Variations
BOF4-S with outlier preservation computes Lloyd-optimal 4-bit block-wise codebooks and beats NF4 and AF4 on quantization error and perplexity.