Imagen de Portada

Improving Block-Wise LLM Quantization by 4-bit Block-Wise Optimal Float (BOF4): Analysis and Variations

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Blumenberg, Patrick, Graave, Thomas, Fingscheidt, Tim
Formato:	Preprint
Publicado:	2025
Materias:	Machine Learning Computation and Language
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

EBFT: Effective and Block-Wise Fine-Tuning for Sparse LLMs
por: Guo, Song, et al.
Publicado: (2024)

LoSA: Locality Aware Sparse Attention for Block-Wise Diffusion Language Models
por: Xi, Haocheng, et al.
Publicado: (2026)

Athena: Efficient Block-Wise Post-Training Quantization for Large Language Models Using Second-Order Matrix Derivative Information
por: Wang, Yanshu, et al.
Publicado: (2024)

Block-Wise Differentiable Sinkhorn Attention: Tail-Refinement Gradients with a Gap-Aware Dustbin Bridge
por: Forde, Dylan
Publicado: (2026)

Block Rotation is All You Need for MXFP4 Quantization
por: Shao, Yuantian, et al.
Publicado: (2025)

Engineering of Hallucination in Generative AI: It's not a Bug, it's a Feature
por: Fingscheidt, Tim, et al.
Publicado: (2026)

BlockDialect: Block-wise Fine-grained Mixed Format Quantization for Energy-Efficient LLM Inference
por: Jang, Wonsuk, et al.
Publicado: (2025)

AAAC: Activation-Aware Adaptive Codebooks for 4-bit LLM Weight Quantization
por: IslamBouli, Beshr, et al.
Publicado: (2026)

LPCD: Unified Framework from Layer-Wise to Submodule Quantization
por: Ichikawa, Yuma, et al.
Publicado: (2025)

Four Over Six: More Accurate NVFP4 Quantization with Adaptive Block Scaling
por: Cook, Jack, et al.
Publicado: (2025)

SmoothRot: Combining Channel-Wise Scaling and Rotation for Quantization-Friendly LLMs
por: Czakó, Patrik, et al.
Publicado: (2025)

KVTuner: Sensitivity-Aware Layer-Wise Mixed-Precision KV Cache Quantization for Efficient and Nearly Lossless LLM Inference
por: Li, Xing, et al.
Publicado: (2025)

CBQ: Cross-Block Quantization for Large Language Models
por: Ding, Xin, et al.
Publicado: (2023)

Multi-task Learning for Heterogeneous Multi-source Block-Wise Missing Data
por: Sui, Yang, et al.
Publicado: (2025)

Adaptive Layer Selection for Layer-Wise Token Pruning in LLM Inference
por: Taniguchi, Rei, et al.
Publicado: (2026)

LO-BCQ: Block Clustered Quantization for 4-bit (W4A4) LLM Inference
por: Elangovan, Reena, et al.
Publicado: (2025)

LoPRo: Enhancing Low-Rank Quantization via Permuted Block-Wise Rotation
por: Gu, Hongyaoxing, et al.
Publicado: (2026)

CHESS: Optimizing LLM Inference via Channel-Wise Thresholding and Selective Sparsification
por: He, Junhui, et al.
Publicado: (2024)

Interactions Across Blocks in Post-Training Quantization of Large Language Models
por: Shabanovi, Khasmamad, et al.
Publicado: (2024)

Sound Source Separation Using Latent Variational Block-Wise Disentanglement
por: Helwani, Karim, et al.
Publicado: (2024)

BitNet a4.8: 4-bit Activations for 1-bit LLMs
por: Wang, Hongyu, et al.
Publicado: (2024)

ParetoQ: Improving Scaling Laws in Extremely Low-bit LLM Quantization
por: Liu, Zechun, et al.
Publicado: (2025)

OrScale: Orthogonalised Optimization with Layer-Wise Trust-Ratio Scaling
por: Lou, Yuxuan, et al.
Publicado: (2026)

BlockEcho: Retaining Long-Range Dependencies for Imputing Block-Wise Missing Data
por: Han, Qiao, et al.
Publicado: (2024)

Listening to the Wise Few: Select-and-Copy Attention Heads for Multiple-Choice QA
por: Tulchinskii, Eduard, et al.
Publicado: (2024)

AgentPRM: Process Reward Models for LLM Agents via Step-Wise Promise and Progress
por: Xi, Zhiheng, et al.
Publicado: (2025)

The Hidden Signal of Verifier Strictness: Controlling and Improving Step-Wise Verification via Selective Latent Steering
por: Zhou, Yefan, et al.
Publicado: (2026)

MaxPoolBERT: Enhancing BERT Classification via Layer- and Token-Wise Aggregation
por: Behrendt, Maike, et al.
Publicado: (2025)

Depth-Wise Attention (DWAtt): A Layer Fusion Method for Data-Efficient Classification
por: ElNokrashy, Muhammad, et al.
Publicado: (2022)

semi-PD: Towards Efficient LLM Serving via Phase-Wise Disaggregated Computation and Unified Storage
por: Hong, Ke, et al.
Publicado: (2025)

GeoBlock: Inferring Block Granularity from Dependency Geometry in Diffusion Language Models
por: Wan, Lipeng, et al.
Publicado: (2026)

Layer-Wise Quantization: A Pragmatic and Effective Method for Quantizing LLMs Beyond Integer Bit-Levels
por: Dumitru, Razvan-Gabriel, et al.
Publicado: (2024)

MAGE: All-[MASK] Block Already Knows Where to Look in Diffusion LLM
por: Kwon, Omin, et al.
Publicado: (2026)

Scaling Laws for Floating Point Quantization Training
por: Sun, Xingwu, et al.
Publicado: (2025)

Optimizing Mixture of Block Attention
por: Xiao, Guangxuan, et al.
Publicado: (2025)

BitNet v2: Native 4-bit Activations with Hadamard Transformation for 1-bit LLMs
por: Wang, Hongyu, et al.
Publicado: (2025)

Dual Path Attribution: Efficient Attribution for SwiGLU-Transformers through Layer-Wise Target Propagation
por: Jantsch, Lasse Marten, et al.
Publicado: (2026)

AttnLRP: Attention-Aware Layer-Wise Relevance Propagation for Transformers
por: Achtibat, Reduan, et al.
Publicado: (2024)

Geometry-Lite: Interpretable Safety Probing via Layer-Wise Margin Geometry
por: Sim, Woo Seob, et al.
Publicado: (2026)

KVSharer: Efficient Inference via Layer-Wise Dissimilar KV Cache Sharing
por: Yang, Yifei, et al.
Publicado: (2024)