EFQ-Softmax is a low‑bit probability‑generation technique that replaces the traditional exp‑then‑quantize path in Transformer attention. It maps shifted attention scores directly to block‑scaled E2M1 operands using an exponent‑only scale and a single affine rule, allowing the same low‑bit representation to be used for both numerator and denominator updates. Experiments on Qwen3‑8B, Qwen3‑VL‑8B‑Instruct, and WAN2.2‑TI2V‑5B show that EFQ‑Softmax maintains or improves model quality while reducing vector‑stage latency by about 40% on the A5 vector unit.
By Haohui Han (Xi'an Jiaotong University), Yuming Wan (Huawei Technologies Co., Ltd), Hongni Wang (Shandong University of Finance and Economics), Pengcheng Xie (Huawei Technologies Co., Ltd), Xiaodong Yan (Xi'an Jiaotong University), Runqi You (Xi'an Jiaotong University), Wencong Zhang (Xi'an Jiaotong University)
arXiv:2609.15810v1 Announce Type: new
Abstract: Diffusion Transformers deliver state-of-the-art video generation, but their long spatiotemporal sequences make attention the dominant deployment cost,...
By Xingyang Li, Dongyun Zou, Shining Zhang, Jiacheng Chen, Haocheng Xi, Lvmin Zhang, Jun-Yan Zhu, Song Han, Zhekai Zhang, Yujun Lin, Muyang Li
arXiv:2607. 04302v1 Announce Type: cross Abstract: We present HiFA4, a post-training operator-level design that executes both QK^T and PV in FlashAttention as 4-bit HIF4 Cube GEMMs for LLM inference on Ascend NPUs, while maintaining the online softmax state in FP16.
By Hui Dong, Yanzhao Li, Jie Gao, Chunlu Li, Zhiyuan Zhang, Yupeng Sun, Zhenyuan Chen, Zhiqiang Zou
arXiv:2607. 24377v1 Announce Type: cross Abstract: The quadratic cost of attention is a major bottleneck in diffusion-based video generation models.
By Jianlin Yu, Jing Lin, Linghui Kong, Aiyue Chen, Weiyi Sun, Chenyu Zeng, Wangli Lan, Jinxi Li, Zhuo Zheng, Ziyang Yue, Danning Ke, Fei Yi, Tianchi Hu, Yuan Ding, Yiwu Yao, Junsong Wang
The paper presents a 4‑bit quantization recipe, Minima: NVFP4 W4A4, that fully quantizes all linear layers—including the Gated DeltaNet (GDN) recurrent blocks—of the 27‑billion‑parameter Qwen3.8 LLM. Across a suite of benchmarks (perplexity, MMLU‑Pro, GSM8K, AIME'25, GPQA‑Diamond, LiveCodeBench, and RULER retrieval), the quantized model matches BF16 performance within seed noise while being 17.5 GiB in size and 14–19 % faster at prefill. The authors attribute this success to four mechanisms: block‑scaling of residuals, robust gate projections, the delta‑rule recurrence’s noise‑plateau behavior, and the per‑token quantization cost’s dilution over long contexts.
By Sergii Kozyrev, Davyd Maiboroda
arXiv:2606. 04620v1 Announce Type: cross Abstract: LLMs have become the state-of-the-art algorithms for solving NLP tasks.
By Pasindu Wickramasinghe, Achyuta Muthuvelan, Rachmad Vidya Wicaksana Putra, Minghao Shao, Muhammad Shafique
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
The paper investigates low‑bit quantization for Multimodal Large Language Models (MLLMs), showing that MXFP8 retains near‑lossless performance while 4‑bit formats like MXFP4 and HiF4 cause significant degradation. It identifies activation quantization as the main source of this loss and introduces Residual Fallback Quantization (RFQ), a lightweight framework that adds a quantized residual pathway to improve activation fidelity without architectural changes. Experiments on Wan2.2 and Qwen3‑VL demonstrate that RFQ recovers much of the performance gap to BF16 baselines across generation and reasoning tasks.
By Tanzila Rahman, Mehran Taghian Jazi, Yunke Peng, Zhuang Ma, Anandharaju Durai Raju, Yao Wang, Xing Huang, Hei Yi Mak, Shadan Golestan, Hoang Le, Yonghan Dong, Wei Guo, Yaoyuan Wang
Multi-modal large language models (MLLMs) with impressive perception and understanding capabilities will enable a range of downstream applications. However, their substantial computational and memory...
The paper introduces CanonQ, a unified quantization‑aware training framework that separates source canonicalization from task‑aware adaptation to enable extreme low‑bit LLM compression. By applying fixed rotations and energy normalization, CanonQ maps heterogeneous tensors to canonical coordinates, allowing frozen Gaussian‑reference codebooks to be reused across layers and models. Joint training then adapts the network to the coupled errors of weight, activation, and KV cache quantization, achieving significant performance gains—up to 14.28× lower perplexity and 57.9% higher zero‑shot accuracy on LLaMA models, and substantial improvements on instruction‑tuned MobileLLM‑Pro‑1B.
"whyItMatters":"The framework demonstrates that carefully designed canonicalization and joint training can dramatically improve the efficiency and accuracy of low‑bit quantized large language models, offering a practical path to deploy powerful LLMs with reduced computational and memory footprints."
By Kai Yi, Tarek Elgamal, Sruthikesh Surineni, Vignesh Vivekraja, Soumyadeep Ghosh, Steven Li
arXiv:2608. 05499v1 Announce Type: cross Abstract: Modern deep neural networks achieve strong performance, but their scale makes them costly and slow, especially on resource-constrained edge devices.
By Sadegh Jafari, Mohiuddin Bilwal, Fan Zhou, Brian Gelder, Ali Jannesari
arXiv:2611.07496v1 Announce Type: cross
Abstract: Attention is a core building block of frontier models and often dominates compute-bound prefill at long context. FP8 matrix instructions offer twice...
By Quentin Anthony, Ganesh Bikshandi, Praneeth Medepalli, Stephen Ebert