The paper proposes two new 4‑bit quantization schemes for AdamW’s optimizer states, focusing on the rounding space of the second moment. Zero‑Inclusive Preconditioner‑space Stochastic Rounding (ZIP‑SR) keeps zero in the codebook and applies stochastic rounding in preconditioner space, while Zero‑Excluding EDEN (ZE‑EDEN) removes zero from the codebook and rescales the quantized block to reduce distortion. Experiments on GPT‑ and Llama‑style models from 130 M to 2.7 B parameters show that both methods close the validation‑loss gap to 32‑bit AdamW, achieving up to a 70 % reduction in the gap and outperforming TorchAO in fine‑tuning tasks.
By Hanyang Li, Shao Tang, Daniel Thomas Braithwaite, Gregory Dexter, Leonardo Neves, Aman Gupta, Hiroto Udagawa, Abhishek Shivanna, Daniel Silva, Rohan Ramanath
arXiv:2608.21952v1 Announce Type: new
Abstract: Recent advances in sequence modeling have highlighted Mamba as a state space architecture offering efficient long-range dependency modeling and providi...
By Hyunwoo Kim, Byoungchan Ko, Minseok Kang, Minwoo Kim, Dongjin Lee, Jaehoon Lee, Sungroh Yoon, Dahuin Jung
arXiv:2608. 08081v1 Announce Type: cross Abstract: Large mixture-of-experts (MoE) language models with 26--120 billion parameters exceed the memory capacity of consumer devices through three simultaneous pressures: resident weight matrices, key-value (KV) cache state that grows linearly with context, and dozens of expert sublayers that must be paged on demand.
By Anthony. Lui, Mohamed. Elsaied, N. P. Savani
arXiv:2609.38169v1 Announce Type: cross
Abstract: Linear attention replaces growing KV caches with fixed-size recurrent states, yet these persistent states can become a substantial memory bottleneck...
By Bingchen Yao, Haobo Xu, Haokun Lin, Yichen Wu, Ziyu Guo, Renrui Zhang, Zhichao Lu, Zhenan Sun, Ying Wei
arXiv:2605. 24011v2 Announce Type: replace-cross Abstract: Vision-Language-Action (VLA) models exhibit remarkable action generation for embodied intelligence, but their heavy compute make deployment on edge platforms impractical.
By Arash Akbari, Arman Akbari, Masih Eskandar, Qitao Tan, Yixiao Chen, Jingwu Luo, Bertha Pangaribuan, Liyun Zhang, Jennifer Dy, Geng Yuan, Xue Lin, Gaowen Liu, Stratis Ioannidis, Yanzhi Wang
arXiv:2607. 02893v1 Announce Type: new Abstract: Low-bit quantization shrinks language models but treats precision as a single global hyper-parameter: every weight uses the same bit-width.
By Hamish Ogilvy
HBQ: Hierarchical Scaling Block Quantization with Hardware‑Efficiency‑Aware Design for Accurate LLM Inference proposes a new block‑quantization scheme that uses large blocks and low‑overhead significand scaling to balance hardware efficiency and accuracy. The authors demonstrate that larger blocks improve efficiency by amortizing dequantization and accumulation costs, while their SIG scaling compensates for the resulting accuracy loss. Experiments on a 28 nm ASIC accelerator show that HBQ achieves up to 4.6× higher area/energy efficiency than state‑of‑the‑art weight‑only quantization, with 1.5–3.0× speedup and 1.6–3.3× system energy reduction over existing BQ methods.
By Chun-Ting Chen, Dongmin Han, Hangyeol Mun, Jake Hyun, Arnab Raha, Amit Agarwal, Mark Anders, Mohamed Abdelfattah, Jae-sun Seo
arXiv:2606. 07819v1 Announce Type: new Abstract: Recently, the efficiency of Large Language Models (LLMs) deployment has become a critical concern in practical applications.
By Hoang-Loc La, Truong-Thanh Le, Amir Taherkordi, Phuong Hoai Ha
arXiv:2607. 28418v1 Announce Type: cross Abstract: Pruning is a promising approach for improving the efficiency of LLMs.
By Haozhe Hu, Hao Wu, Peiran Yin, Chao Han, Yunpu Ma, Xiaoyu Shen
arXiv:2609.26333v1 Announce Type: new
Abstract: Prefill and decode reward different approaches to quantization: low-precision arithmetic accelerates prompt processing, while compact weights reduce me...
By Andrei Panferov, Maximilian Kleinegger, Sweta Priyadarshi, Tijmen Blankevoort, Dan Alistarh
ActQuant is a mixed‑precision post‑training quantization framework that reduces Vision‑Language‑Action model weights to sub‑4‑bit precision. It first allocates a single bit‑width to each weight matrix based on its contribution to action prediction, then optimizes per‑block scales using action‑aware curvature to focus dynamic range on control‑critical weights. Coupled with the OmniModel.cpp conversion pipeline, ActQuant enables deployment on edge devices, achieving up to 2.5 bits‑per‑weight while preserving over 90 % of baseline performance on both simulation and a real 6‑DoF UR3 robotic arm.
By Arash Akbari, Arman Akbari, Masih Eskandar, Qitao Tan, Yixiao Chen, Jingwu Luo, Bertha Pangaribuan, Liyun Zhang, Jennifer Dy, Geng Yuan, Xue Lin, Gaowen Liu, Stratis Ioannidis, Yanzhi Wang
arXiv:2609.39223v2 Announce Type: new
Abstract: Large language model (LLM) inference is increasingly moving toward lower precision to realize the throughput of hardware accelerators, but aggressive p...
By Weili Xu, Jisen Li, Yuqing Jian, Chenxi Li, Zhizhou Sha, Yifan Yu, Qingyang Wu, Chenfeng Xu, Zhongzhu Zhou, Tianyi Zhang, Ben Athiwaratkun