The paper introduces MiX, a micro‑inverted‑scaling format that replaces shared exponents with shared mantissas to avoid microscaling collapse in low‑bit vision‑language models. An adaptive dual‑format inference framework (MiX‑MX) maps this format to a custom accelerator, replacing multipliers with shifters. Experiments show 4.5‑bit MiX matches or outperforms NVFP4 accuracy while improving area efficiency by 25 % and delivering 2.3–4.5× speedup with 1.4–2.9× energy savings over the Focus accelerator.
By Yuan Liao, Jae-sun Seo
DualQuant introduces a dual‑scale, calibration‑free post‑training quantization framework that jointly optimizes weight‑reconstruction MSE and derives per‑channel activation rescalings, enabling effective 4‑bit weight and activation quantization without calibration data. The method employs a non‑differentiable optimization solved via variable splitting and block coordinate descent, supporting a wide range of formats including MX and NVFP4. Experiments on Llama and Qwen models show DualQuant matches or surpasses calibration‑based SmoothQuant and significantly outperforms other calibration‑free baselines, achieving up to 2.61× memory reduction and 2.46× throughput speedup on an RTX 4090.
By Mahsa Salmani, Sayeh Sharify, Hesham Mostafa
arXiv:2605. 24391v2 Announce Type: replace-cross Abstract: As the demand for deep learning grows, cost reduction through quantization has become essential for both training and inference.
By Dahoon Park, Jahyun Koo, Sangwoo Hwang, Jaeha Kung
arXiv:2608. 15602v1 Announce Type: cross Abstract: While binary quantization theoretically promises extreme compression and acceleration for Large Language Models (LLMs), existing research often overlooks the necessity of specialized hardware kernels, thus failing to unleash the full acceleration potential due to persistent reliance on expensive floating-point arithmetic or runtime dequantization overheads.
By Qingyao Yang, Runming Yang, He Xiao, Wendong Xu, Junyu Chen, Haobo Liu, Chenchen Ding, Ruihan Hu, Yik-Chung Wu, Ngai Wong
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:2607. 08029v1 Announce Type: new Abstract: The emergence of vision language models with fewer than 3 billion parameters has accelerated the implementation of on-device multimodal intelligence.
By Hyeju Shin, Chorwon Kim, Ryangsoo Kim, Hark Yoo, Jaein Kim
ProQuant introduces a progressive quantization‑aware training framework for low‑bit multi‑modal large language models (MLLMs), addressing the performance drop seen with post‑training quantization at low bit‑widths. The method includes a self‑relabeling scheme to align fine‑tuning data with pre‑training distributions and a progressive block activation mechanism that prioritizes multi‑modal understanding, guided by a vision‑sensitive factor to identify critical layers. Experiments on 2B–8B models show ProQuant outperforms existing techniques, achieving up to a 6.6% accuracy gain in W4A8 Qwen3‑VL‑2B‑Instruct with only ~1% drop from full precision.
By Yufei Xue, Yushi Huang, Jiawei Shao, Pingcheng Dong, Yonghao Tan, Shiyao Li, Kwang-Ting Cheng, Jun Zhang
ProQuant introduces a progressive quantization-aware training framework for low-bit multi‑modal large language models (MLLMs). It employs a self‑relabeling scheme to align finetuning data with pretrained distributions and a progressive block activation mechanism that prioritizes multi‑modal understanding, guided by a vision‑sensitive factor. Experiments on 2B‑8B models show significant gains, such as a 6.6% accuracy improvement for W4A8 Qwen3‑VL‑2B‑Instruct with only ~1% drop from full precision.
arXiv:2509. 10334v2 Announce Type: replace-cross Abstract: Vision Transformers (ViTs) have recently achieved strong results in semantic segmentation, yet their deployment on resource-constrained devices remains limited due to their high memory footprint and computational cost.
By Jordan Sassoon, Michal Szczepanski, Martyna Poreba
VisionMX introduces a post‑training microscaling (MX) quantization technique for vision models, addressing three key error sources identified in direct conversion: block‑scale representation, misalignment of small convolutional weight tensors, and underuse of signed codes for nonnegative activations. The method optimizes bounded weight rounding and incorporates a foldable affine correction for activations, improving performance across image classification, object detection, semantic segmentation, and low‑light image enhancement tasks. VisionMX outperforms direct conversion and other post‑training quantization baselines, especially in architectures most sensitive to MX conversion.
By Elad Dror Cohen, Ofir Gordon, Lior Dikstein, Idan Achituve, Hai Victor Habi
arXiv:2605. 26092v4 Announce Type: replace-cross Abstract: The deployment of Large Language Models (LLMs) and Vision Transformers (ViTs) on edge devices is significantly constrained by memory limitations and the critical timing bottlenecks introduced by dense Multiply-Accumulate (MAC) arrays.
By Maoyang Xiang, Tao Luo, Bo Wang
arXiv:2410. 13056v4 Announce Type: replace-cross Abstract: Large Language Models (LLMs) have demonstrated remarkable success across a wide range of language tasks, but their deployment on edge devices remains challenging due to the substantial memory requirements imposed by their large parameter sizes.
By Zihan Chen, Bike Xie, Jundong Li, Cong Shen