TileMix is a tile‑centric mixed‑precision attention kernel that routes score‑tile groups within fused dense attention to either FP16 or INT8 computation, using compact bitmasks to decide precision per tile. By partitioning the attention matrix into hardware‑aligned tiles and updating a shared online‑softmax state, TileMix preserves dense token connectivity without requiring training and supports grouped‑query attention, variable‑length batches, and INT8 key/value caches. Benchmarks on LLaMA, Qwen, and Vicuna show that TileMix restores long‑context quality lost with uniform INT8 and improves prefill throughput over FP16, offering a controllable accuracy‑efficiency trade‑off across model families.
By Hanzhi Zhang, Qiao Zhang, Qinglei Cao, Heng Fan, Yan Huang, Kewei Sha, Yunhe Feng
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:2601. 07475v2 Announce Type: replace-cross Abstract: The emergence of fine-grained numerical formats like NVFP4 presents new opportunities for efficient Large Language Model (LLM) inference.
By Haoqian Meng, Yilun Luo, Yafei Zhao, Wenyuan Liu, Peng Zhang, Xindian Ma
arXiv:2608. 06763v1 Announce Type: new Abstract: Weight quantization for large-language-model inference must balance adaptive reconstruction levels with representations regular enough for efficient GPU execution.
By Xuetian Gao
arXiv:2605. 08692v2 Announce Type: replace Abstract: Post-training weight-only quantization to 4 bits is widely used to reduce the memory and compute costs of large language model inference.
By Beshr IslamBouli, David Jin
arXiv:2607. 23047v1 Announce Type: cross Abstract: Mixed-precision quantization improves the accuracy of post-training quantization by allocating higher bitwidths to sensitive layers, but existing methods solve the allocation for a single fixed memory budget.
By Ashitabh Misra, Madhav Agrawal, Arham Jain, Tarek Abdelzaher
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
arXiv:2602. 01027v2 Announce Type: replace Abstract: Mixed-precision quantization is a promising approach for compressing large language models under tight memory budgets.
By Xin Nie, Haicheng Zhang, Liang Dong, Beining Feng, Jinhong Weng, Guiling Sun
arXiv:2606. 11357v1 Announce Type: cross Abstract: With the growing demand for on-device LLM inference, edge SoCs increasingly integrate NPUs to improve performance and energy efficiency under tight power and thermal budgets.
By Wesley Pang, Gregory Hyegang Jun, Feiyang Liu, Deming Chen
Weight quantization for large-language-model inference must balance adaptive reconstruction levels with representations regular enough for efficient GPU execution. Uniform integers constrain each group to a linear grid.
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
The paper addresses the problem of non‑deterministic outputs from large language models (LLMs) when run on different GPU architectures, caused by floating‑point non‑associativity and hardware‑dependent kernel choices. It proposes a set of fixed‑configuration fused‑upcast GEMM kernels that load 16‑bit weights, upcast to FP32, and perform IEEE‑754 compliant reductions in a problem‑shape‑dependent order, ensuring identical linear‑layer outputs across NVIDIA Ampere, Ada, and Hopper GPUs. The new approach achieves 1.17–3.1× faster end‑to‑end performance than existing solutions and halves weight‑memory traffic while maintaining cross‑architecture reproducibility.
By Liam Cooper, Shinnung Jeong, Hyeran Jeon, Jeffrey Young, Hyesoon Kim