Related papers: FlowKV: A Disaggregated Inference Framework with L…

The serving paradigm of large language models (LLMs) is rapidly shifting towards complex multi-agent workflows where specialized agents collaborate over massive shared contexts. While Low-Rank Adaptation (LoRA) enables the efficient…

Key-value (KV) cache memory management is the primary bottleneck limiting throughput and cost-efficiency in large-scale GPU inference serving. Current systems suffer from three compounding inefficiencies: (1) the absence of unified KV cache…

The expanding context windows in large language models (LLMs) have greatly enhanced their capabilities in various applications, but they also introduce significant challenges in maintaining low latency, particularly in Time to First Token…

Vision-Language-Action (VLA) models offer a unified framework for robotic perception and control, but their ability to scale to real-world, long-horizon tasks is limited by the high computational cost of attention and the large memory…

Graph neural networks (GNNs) have recently exploded in popularity thanks to their broad applicability to graph-related problems such as quantum chemistry, drug discovery, and high energy physics. However, meeting demand for novel GNN models…

As large language models (LLMs) continue to scale, the memory footprint of key-value (KV) caches during inference has become a significant bottleneck. Existing approaches primarily focus on compressing KV caches within a single prompt or…

Large language models (LLMs) rely on key-value cache (KV cache) to accelerate decoding by reducing redundant computations. However, the KV cache memory usage grows substantially with longer text sequences, posing challenges for efficient…

Diffusion-based large language models (dLLMs) have emerged as a promising alternative to autoregressive (AR) LLMs, leveraging denoising-based generation to enable inherent parallelism. Even more and more open-sourced dLLM models emerge, yet…

Efficient LLM inference is critical for real-world applications, especially within heterogeneous GPU clusters commonly found in organizations and on-premise datacenters as GPU architecture rapidly evolves. Current disaggregated prefill…

Contemporary systems serving large language models (LLMs) have adopted prefill-decode disaggregation to better load-balance between the compute-bound prefill phase and the memory-bound decode phase. Under this design, prefill workers…

Key-value (KV) caching is widely used to accelerate transformer inference, but its memory cost grows linearly with input length, limiting long-context deployment. Existing token eviction methods reduce memory by discarding less important…

Large Reasoning Models (LRMs) are becoming integral to many AI inference systems, enhancing their capabilities with advanced reasoning. However, deploying these models in production environments presents a significant QoS challenge: the…

We introduce the Block Transformer which adopts hierarchical global-to-local modeling to autoregressive transformers to mitigate the inference bottlenecks associated with self-attention. Self-attention requires the key-value (KV) cache of…

Large Language Models (LLMs) have been widely adopted to process long-context tasks. However, the large memory overhead of the key-value (KV) cache poses significant challenges in long-context scenarios. Existing training-free KV cache…

KV cache quantization reduces the memory cost of long-context LLM inference, but introduces approximation error that is typically validated only empirically. Existing systems rely on average-case robustness, with no mechanism to detect or…

In this paper, we introduce LightVLM, a simple but effective method that can be seamlessly deployed upon existing Vision-Language Models (VLMs) to greatly accelerate the inference process in a training-free manner. We divide the inference…

Long-context inference in large language models (LLMs) is increasingly constrained by the KV cache bottleneck: memory usage grows linearly with sequence length, while attention computation scales quadratically. Existing approaches address…

In long-context large language model (LLM) inference, the prefill stage dominates computation due to self-attention over the complete input context. Sparse attention significantly reduces self-attention computation by limiting each token's…

Cache fusion accelerates generation process of LLMs equipped with RAG through KV caching and selective token recomputation, thereby reducing computational costs and improving efficiency. However, existing methods primarily rely on local…

Video generation is pivotal to digital media creation, and recent advances in autoregressive video generation have markedly enhanced the efficiency of real-time video synthesis. However, existing approaches generally rely on heuristic KV…