Related papers: Token-Budget-Aware Pool Routing for Cost-Efficient…

Recent breakthroughs in Large-scale language models (LLMs) have demonstrated impressive performance on various tasks. The immense sizes of LLMs have led to very high resource demand and cost for running the models. Though the models are…

Modern Large Language Model serving system batches multiple requests to achieve high throughput, while batching attention operations is challenging, rendering memory bandwidth a critical bottleneck. The community relies on high-end GPUs…

WebGPU's security-focused design imposes per-operation validation that compounds across the many small dispatches in neural network inference, yet the true cost of this overhead is poorly characterized. We present a systematic…

Vision-language models (VLMs) excel at image understanding tasks, but the large number of visual tokens imposes significant computational costs, hindering deployment on mobile devices. Many pruning methods rely solely on token importance…

Large Language Models (LLMs) rely on the Key-Value (KV) Cache to store token history, enabling efficient decoding of tokens. As the KV-Cache grows, it becomes a major memory and computation bottleneck. However, there is an opportunity to…

Optimizing the Key-Value (KV) cache of the Large Language Model (LLM) has been considered critical to saving the cost of inference. Most of the existing KV-cache compression algorithms attempted to sparsify the sequence of tokens by taking…

Vision-Language Models (VLMs) demand substantial computational resources during inference, largely due to the extensive visual input tokens for representing visual information. Previous studies have noted that visual tokens tend to receive…

The rapid growth of large language model (LLM) services imposes increasing demands on distributed GPU inference infrastructure. Most existing scheduling systems follow a reactive paradigm, relying solely on the current system state to make…

Reasoning LLMs often spend substantial tokens on long intermediate reasoning traces (e.g., chain-of-thought) when solving new problems. We propose to summarize and store reusable reasoning skills distilled from extensive deliberation and…

Large language models (LLMs) increasingly rely on long-context processing, but expanding context windows introduces substantial computational and financial costs. Existing context reduction approaches, including retrieval and memory…

Large language model (LLM) applications are blindfolded to the infrastructure underneath and generate tokens autoregressively, indifferent to the system load, thus risking inferencing latency inflation and poor user experience. Our…

The key-value (KV) cache is a foundational optimization in Transformer-based large language models (LLMs), eliminating redundant recomputation of past token representations during autoregressive generation. However, its memory footprint…

AI-enabled systems are subjected to various types of runtime uncertainties, ranging from dynamic workloads, resource requirements, model drift, etc. These uncertainties have a big impact on the overall Quality of Service (QoS). This is…

In real-world Tool-Integrated Reasoning (TIR) scenarios, where LLMs interleave reasoning with external tool calls, a major source of inefficiency is that the toolcalls create pauses between LLM requests and cause KV-Cache eviction, forcing…

To address the growing demand for on-device LLM inference in resource-constrained environments, hybrid language models (HLM) have emerged, combining lightweight local models with powerful cloud-based LLMs. Recent studies on HLM have…

Selecting the appropriate model at inference time -- the routing problem -- requires jointly optimizing output quality, cost, latency, and governance constraints. Existing approaches delegate this decision to LLM-based classifiers or…

In this study, we identify the inefficient attention phenomena in Large Vision-Language Models (LVLMs), notably within prominent models like LLaVA-1.5, QwenVL-Chat and Video-LLaVA. We find out that the attention computation over visual…

Generative Recommender (GR) inference places embedding hot caches (EMB) and KV caches in direct competition for limited GPU HBM: allocating more memory to one improves its efficiency but degrades the other. Existing systems optimize them in…

Given the quadratic complexity of attention, KV cache eviction is vital to accelerate model inference. Current KV cache eviction methods typically rely on instantaneous heuristic metrics, implicitly assuming that score magnitudes are…

The widespread of Large Language Models (LLMs) marks a significant milestone in generative AI. Nevertheless, the increasing context length and batch size in offline LLM inference escalate the memory requirement of the key-value (KV) cache,…