Cover Image

MX-SAFE: Versatile Inference- and Training-Proof Microscaling Format with On-the-Fly Exponent and Mantissa Bit Allocation

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Park, Dahoon, Koo, Jahyun, Hwang, Sangwoo, Kung, Jaeha
Format:	Preprint
Published:	2026
Subjects:	Hardware Architecture Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

OPAL: Outlier-Preserved Microscaling Quantization Accelerator for Generative Large Language Models
by: Koo, Jahyun, et al.
Published: (2024)

MX+: Pushing the Limits of Microscaling Formats for Efficient Large Language Model Serving
by: Lee, Jungi, et al.
Published: (2025)

Dissecting and Re-architecting 3D NAND Flash PIM Arrays for Efficient Single-Batch Token Generation in LLMs
by: Jang, Yongjoo, et al.
Published: (2025)

VMXDOTP: A RISC-V Vector ISA Extension for Efficient Microscaling (MX) Format Acceleration
by: Wipfli, Max, et al.
Published: (2026)

SageAttention3: Microscaling FP4 Attention for Inference and An Exploration of 8-Bit Training
by: Zhang, Jintao, et al.
Published: (2025)

Efficient Precision-Scalable Hardware for Microscaling (MX) Processing in Robotics Learning
by: Cuyckens, Stef, et al.
Published: (2025)

MXDOTP: A RISC-V ISA Extension for Enabling Microscaling (MX) Floating-Point Dot Products
by: İslamoğlu, Gamze, et al.
Published: (2025)

A Dataflow Compiler for Efficient LLM Inference using Custom Microscaling Formats
by: Cheng, Jianyi, et al.
Published: (2023)

Characterization and Mitigation of Training Instabilities in Microscaling Formats
by: Su, Huangyuan, et al.
Published: (2025)

Efficient Reprogramming of Memristive Crossbars for DNNs: Weight Sorting and Bit Stucking
by: Farias, Matheus, et al.
Published: (2024)

All-rounder: A Flexible AI Accelerator with Diverse Data Format Support and Morphable Structure for Multi-DNN Processing
by: Noh, Seock-Hwan, et al.
Published: (2023)

MicroScopiQ: Accelerating Foundational Models through Outlier-Aware Microscaling Quantization
by: Ramachandran, Akshat, et al.
Published: (2024)

Jack Unit: An Area- and Energy-Efficient Multiply-Accumulate (MAC) Unit Supporting Diverse Data Formats
by: Noh, Seock-Hwan, et al.
Published: (2025)

Towards Cheaper Inference in Deep Networks with Lower Bit-Width Accumulators
by: Blumenfeld, Yaniv, et al.
Published: (2024)

MixDiT: Accelerating Image Diffusion Transformer Inference with Mixed-Precision MX Quantization
by: Kim, Daeun, et al.
Published: (2025)

PIMphony: Overcoming Bandwidth and Capacity Inefficiency in PIM-based Long-Context LLM Inference System
by: Kwon, Hyucksung, et al.
Published: (2024)

BF-IMNA: A Bit Fluid In-Memory Neural Architecture for Neural Network Acceleration
by: Rakka, Mariam, et al.
Published: (2024)

Row-Column Hybrid Grouping for Fault-Resilient Multi-Bit Weight Representation on IMC Arrays
by: Jeon, Kang Eun, et al.
Published: (2025)

Precision-Scalable Microscaling Datapaths with Optimized Reduction Tree for Efficient NPU Integration
by: Cuyckens, Stef, et al.
Published: (2025)

Panacea: Novel DNN Accelerator using Accuracy-Preserving Asymmetric Quantization and Energy-Saving Bit-Slice Sparsity
by: Kam, Dongyun, et al.
Published: (2024)

BitParticle: Partializing Sparse Dual-Factors to Build Quasi-Synchronizing MAC Arrays for Energy-efficient DNNs
by: Qiaoyuan, Feilong, et al.
Published: (2025)

ODMA: On-Demand Memory Allocation Strategy for LLM Serving on LPDDR-Class Accelerators
by: Zou, Guoqiang, et al.
Published: (2025)

Rethinking LLM Inference Bottlenecks: Insights from Latent Attention and Mixture-of-Experts
by: Yun, Sungmin, et al.
Published: (2025)

A 10.60 $μ$W 150 GOPS Mixed-Bit-Width Sparse CNN Accelerator for Life-Threatening Ventricular Arrhythmia Detection
by: Qin, Yifan, et al.
Published: (2024)

HiFloat4 Format for Language Model Inference
by: Luo, Yuanyong, et al.
Published: (2026)

Refining Datapath for Microscaling ViTs
by: Xiao, Can, et al.
Published: (2025)

Pre-gated MoE: An Algorithm-System Co-Design for Fast and Scalable Mixture-of-Expert Inference
by: Hwang, Ranggi, et al.
Published: (2023)

HiKonv: Maximizing the Throughput of Quantized Convolution With Novel Bit-wise Management and Computation
by: Chen, Yao, et al.
Published: (2022)

Scalable Processing-Near-Memory for 1M-Token LLM Inference: CXL-Enabled KV-Cache Management Beyond GPU Limits
by: Kim, Dowon, et al.
Published: (2025)

M2XFP: A Metadata-Augmented Microscaling Data Format for Efficient Low-bit Quantization
by: Hu, Weiming, et al.
Published: (2026)

Pushing the Limits of BFP on Narrow Precision LLM Inference
by: Wang, Hui, et al.
Published: (2025)

Comparative Characterization of KV Cache Management Strategies for LLM Inference
by: Mamo, Oteo, et al.
Published: (2026)

ELSA: An ELastic SNN Inference Architecture for Efficient Neuromorphic Computing
by: You, Kang, et al.
Published: (2026)

Leveraging Compute-in-Memory for Efficient Generative Model Inference in TPUs
by: Zhu, Zhantong, et al.
Published: (2025)

Dynamic Gradient Sparse Update for Edge Training
by: Li, I-Hsuan, et al.
Published: (2025)

KANtize: Exploring Low-bit Quantization of Kolmogorov-Arnold Networks for Efficient Inference
by: Errabii, Sohaib, et al.
Published: (2026)

Sangam: Chiplet-Based DRAM-PIM Accelerator with CXL Integration for LLM Inferencing
by: Kiyawat, Khyati, et al.
Published: (2025)

FlexNeRFer: A Multi-Dataflow, Adaptive Sparsity-Aware Accelerator for On-Device NeRF Rendering
by: Noh, Seock-Hwan, et al.
Published: (2025)

Enhancing LUT-based Deep Neural Networks Inference through Architecture and Connectivity Optimization
by: Lou, Binglei, et al.
Published: (2026)

Embedded FPGA Acceleration of Brain-Like Neural Networks: Online Learning to Scalable Inference
by: Hafiz, Muhammad Ihsan Al, et al.
Published: (2025)