UniPrefill is a prefill‑acceleration framework that works with virtually any model architecture, directly speeding up token‑level computation. It is implemented as a continuous‑batching operator and extends vLLM’s scheduling to support prefill‑decode co‑processing and tensor parallelism. The method delivers up to a 2.1× speedup in Time‑to‑First‑Token, with gains growing as concurrent requests increase.
By Qihang Fan, Huaibo Huang, Zhiying Wu, Bingning Wang, Ran He
The paper introduces a quantization technique for tabular foundation models that focuses on converting queries, keys, and values to FP8 and employing explicit FP8 matrix multiplication to accelerate attention calculations. It emphasizes aligning quantization errors between training and test rows to avoid accuracy loss, and demonstrates up to 1.7× speedup over 16‑bit kernels with no significant accuracy degradation on TabPFN‑v3 and TabICLv2 across TabArena and BeyondArena.
By Jonas M. K\"ubler, Benjamin J\"ager, Klemens Fl\"oge, Noah Hollmann, Frank Hutter
arXiv:2608.30320v1 Announce Type: new
Abstract: We describe the architecture and ablations of Qwen3.8-Flash-Next, a sparse mixture-of-experts model with 125B parameters, 6B activated per token, and a...
By Zihan Qiu, Zekun Wang, Xiao Li, Yanpeng Li, Yang Xu, Yixuan Wang, Huaqing Zhang, Rui Men, Bochao Mao, Chengruidong Zhang, Fan Zhou, Hao Luo, Haofeng Huang, Haoran Lian, Haoyan Huang, Hongqing Chen, Jianwei Zhang, Jing Xu, Junjie Wang, Langshi Chen, Liangyu Wang, Linlang Jiang, Man Yuan, Minmin Sun, Peng Jin, Siqi Zhang, Siyu Wang, Xingzhang Ren, Yakai Wang, Yi Zhang, Yiming Dong, Yizhong Cao, Yubo Ma, Yunfei Mao, Bo Zheng, Dayiheng Liu
arXiv:2606. 13392v1 Announce Type: new Abstract: Ultra-long-context capability is becoming indispensable for frontier LLMs: agentic workflows, repository-scale code reasoning, and persistent memory all require the model to jointly attend over hundreds of thousands to millions of tokens, yet the quadratic cost of softmax attention makes this untenable at deployment scale.
By Xunhao Lai, Weiqi Xu, Yufeng Yang, Qiaorui Chen, Yang Xu, Lunbin Zeng, Xiaolong Li, Haohai Sun, Haichao Zhu, Vito Zhang, Pengyu Zhao
arXiv:2610.07940v1 Announce Type: cross
Abstract: Looped language models apply the same stack of layers T times to each token, which deepens the model without adding parameters but multiplies its key...
By Yuhan Chen, Siyuan Zhang, Nan Wang, Feiyang Kang, Ruoxi Jia
arXiv:2512.20636v2 Announce Type: replace-cross
Abstract: Many self-attention sublayers in large language models (LLMs) can be removed with little to no loss. We attribute this to the Attention Suppr...
By Dhananjay Saikumar, Blesson Varghese
arXiv:2608. 06628v1 Announce Type: new Abstract: Diffusion language models (dLLMs) offer a promising alternative to autoregressive models by accelerating inference through parallel decoding.
By Jinha Kim, Younghun Roh, Jaeyeon Kim
Pruning large pre-trained transformer-based ASR models such as OpenAI's Whisper has seen great adoption, as pruning the decoder led to significant end-to-end transcription speedups. For instance, the...
MONA is a new optimizer that extends the Muon optimizer by adding a Nesterov‑style acceleration term derived from an exponential moving average of gradient differences. The paper provides a convergence analysis showing that this term offers curvature‑aware corrections while maintaining Muon’s spectral‑norm regularization. Empirical results demonstrate that MONA outperforms both Muon and AdamW on Mixture‑of‑Experts pretraining across models ranging from 1 B to 68 B parameters, and achieves state‑of‑the‑art performance on downstream benchmarks after fine‑tuning the largest model.
By Jiacheng Li, Jianchao Tan, Hongtao Xu, Jiaqi Zhang, Yifan Lu, Yerui Sun, Yuchen Xie, Xunliang Cai
arXiv:2607. 07953v1 Announce Type: cross Abstract: Self-attention lets each token retrieve information from the full context, but its quadratic cost in sequence length limits training and inference at long context.
By Tommaso Cerruti, Tim Rieder, George Rowlands, Lingfeng Jin, Imanol Schlag
arXiv:2509. 10406v4 Announce Type: replace Abstract: Pretraining transformers on long sequences (entire code repositories, collections of related documents) is bottlenecked by quadratic attention costs.
By Rupert Mitchell, Kristian Kersting
LoopICL is a transformer architecture that loops a single block to address tabular tasks. It separates parameter count from computational depth by using a cell stream for per‑cell features and a row stream for in‑context examples, refined via within‑column and cross‑column attention. During pre‑training, varying loop counts and a learned exit‑gate allow the model to adjust inference depth at test time, achieving competitive performance with TabICLv2 while using about 90% fewer parameters.
By Amir Rezaei Balef, Katharina Eggensperger