arXiv Computation and Language

Efficient Iterative Retrieval with Heterogeneous Batching

Orthrus is a serving system that performs heterogeneous batching of embedding and generative models within a single inference loop. It uses chunked embedding with incremental pooling and workload‑aware batch composition to unify conflicting computational patterns. Experiments on four A100 GPUs show that Orthrus improves throughput by 1.28×–4.52× and reduces p99 latency by up to 55.8% compared to baseline deployments.

arXiv Machine Learning
Oct 9

QUILT: Rethinking Sparse-Attention Prefill through Shared Query Execution

The paper introduces QUILT, a new sparse‑attention execution mechanism that jointly processes neighboring queries to reuse shared key‑value (KV) entries, thereby reducing redundant memory traffic and computation. QUILT employs Shift‑and‑Compare Set Decomposition (SCSD) to convert irregular set operations into regular data‑parallel primitives and pipelines this with attention computation to hide overhead. The approach includes cascaded sharing across multiple granularities and a tile‑aware strategy that balances sharing with hardware utilization, achieving up to 55.1% lower kernel latency and 36.8% faster time‑to‑first‑token on LongBench benchmarks with negligible accuracy loss.

By Zhenduo Zhao, Qihui Zhou, Mingcong Song, Zhiyi Chen, Chuangguan Ye, Fengfan Hou, Zequn Gong, Jing Li, Hongjie Si, Guoping Long
arXiv Machine Learning
Sep 21

Hybrid GPU-CPU Retrieval for Personalized Search at Ultra-Large Scale

arXiv:2609.21281v1 Announce Type: cross Abstract: Embedding-based retrieval on user-generated content at the trillion-document scale exposes a sharp conflict between two production demands: deep, exp...

By Hao Fu, Jichao Sun, Baiting Zhu, Qiaoling Liu, Yan Shi, Cheng Lu, Liu Liu, Yubo Wang, Xin Yao, Xiangyu Niu, Xu Dong, Wenhan Lyu, Chiyao Shen, Yinjie Huang, Minglei Chen, Shuai Ding, Li Fan, Xiao Kong
arXiv Machine Learning
Jul 22

Vectorizing the Trie: Efficient Constrained Decoding for LLM-based Generative Retrieval on Accelerators

arXiv:2602. 22647v2 Announce Type: replace-cross Abstract: Generative retrieval has emerged as a powerful paradigm for LLM-based recommendation.

By Zhengyang Su, Isay Katsman, Yueqi Wang, Ruining He, Lukasz Heldt, Raghunandan Keshavan, Shao-Chuan Wang, Xinyang Yi, Mingyan Gao, Onkar Dalal, Lichan Hong, Ed Chi, Ningren Han
arXiv Machine Learning
Oct 12

STAPLE: Static Sparse Upcycling of Pretrained Checkpoints via Embedding Modules

STAPLE is a method that integrates static sparse memory into pretrained language models by adding token‑indexed embeddings as a residual pathway while preserving the original feed‑forward network. On the OLMo‑2 1B and 7B models, STAPLE outperforms dense training and other static sparse baselines, improving benchmark scores by up to 4.27 points and achieving faster decoding with less GPU memory and energy usage. The approach demonstrates that large pretrained models can be expanded efficiently without costly sparse pretraining from scratch.

By Ranajoy Sadhukhan, Euijin Hong, Sheng Cao, Zechun Liu, Shuowei Jin, Sha Li, Changlong Yu, Xin Liu, Qin Lu, Zhenyu Shi, Priyanka Nigam, Beidi Chen
arXiv Machine Learning
Jun 9

BlendServe: Optimizing Offline Inference for Auto-regressive Large Models with Resource-aware Batching

arXiv:2411. 16102v2 Announce Type: replace Abstract: Offline batch inference, which leverages the flexibility of request batching to achieve higher throughput and lower costs, is becoming more popular for latency-insensitive applications.

By Yilong Zhao, Shuo Yang, Kan Zhu, Lianmin Zheng, Baris Kasikci, Yang Zhou, Jiarong Xing, Ion Stoica
arXiv AI
Oct 12

Adaptive DHiSparse: Speculative Decoding over a Load-Adaptive KV Hierarchy

Adaptive DHiSparse extends the HiSparse framework by enabling speculative decoding over a hierarchical sparse key‑value (KV) store and dynamically adapting memory residency to load. It protects the essential working set, retains extra history when GPU memory is available, and retrieves the remainder via exact fused retrieval, integrating draft‑model techniques such as DSpark and DFlash2. In benchmarks, it achieves up to 1.85× higher output rates than resident serving while reducing host fetches dramatically under low load.

By Naji Alazhar, Wassel Alazhar
arXiv Machine Learning
Sep 17

GroupKV: Hierarchical KV Cache Management for Long-Context Diffusion LLM Inference

GroupKV is a lightweight hierarchical KV cache management system designed for long‑context diffusion large language model (dLLM) inference. It partitions the context into contiguous groups and uses coarse‑to‑fine sparse selection, cross‑layer consistency for predictive prefetching, and a staleness correction mechanism to keep the cache coherent amid dynamic KV updates. The approach also incorporates streaming prefill to lower peak memory usage, achieving up to 48× longer serviceable context, 3.73× faster inference in offload‑based settings, and competitive task accuracy.

By Jinhao Wang, Zhexin Hu, Kangjie Zhou, Xin Zhou, Fangfang Liu