It consists of two components, i.e., a cross-decoder stacked upon a self-decoder .

Efficient state restoration in multi-turn conversations with large language models (LLMs) remains a critical challenge, primarily due to the overhead of recomputing or loading full key-value (KV) caches for all historical tokens. To address this, existing approaches compress KV caches across adjacent layers with highly similar attention patterns. However, these methods often apply a fixed compression scheme across all conversations, selecting the same layer pairs for compression without considering conversation-specific attention dynamics. This static strategy overlooks variability in attention pattern similarity across different conversations, which can lead to noticeable accuracy degradation. We present Krul, a multi-turn LLM inference system that enables accurate and efficient KV cache restoration. Krul dynamically selects compression strategies based on attention similarity across layer pairs and uses a recomputation-loading pipeline to restore the KV cache. It introduces three key innovations: 1) a preemptive compression strategy selector to preserve critical context for future conversation turns and selects a customized strategy for the conversation; 2) a token-wise heterogeneous attention similarity estimator to mitigate the attention similarity computation and storage overhead during model generation; 3) a bubble-free restoration scheduler to reduce potential bubbles brought by the imbalance of recomputing and loading stream due to compressed KV caches. Empirical evaluations on real-world tasks demonstrate that Krul achieves a 1.5x-2.68x reduction in time-to-first-token (TTFT) and a 1.33x-2.35x reduction in KV cache storage compared to state-of-the-art methods without compromising generation quality.

Multiple Low-Rank Adapters (Multi-LoRAs) are gaining popularity for task-specific Large Language Model (LLM) applications. For multi-LoRA serving, caching hot KV caches and LoRA adapters in high bandwidth memory of accelerations can improve inference performance. However, existing Multi-LoRA inference systems fail to optimize serving performance like Time-To-First-Toke (TTFT), neglecting usage dependencies when caching LoRAs and KVs. We therefore propose FASTLIBRA, a Multi-LoRA caching system to optimize the serving performance. FASTLIBRA comprises a dependency-aware cache manager and a performance-driven cache swapper. The cache manager maintains the usage dependencies between LoRAs and KV caches during the inference with a unified caching pool. The cache swapper determines the swap-in or out of LoRAs and KV caches based on a unified cost model, when the HBM is idle or busy, respectively. Experimental results show that ELORA reduces the TTFT by 63.4% on average, compared to state-of-the-art works.

Recent large language models (LLMs) face increasing inference latency as input context length and model size continue to grow. In particular, the retrieval-augmented generation (RAG) technique, which enhances LLM responses by incorporating external knowledge, exacerbates this issue by significantly increasing the number of input tokens. This expansion in token length leads to a substantial rise in computational overhead, particularly during the prefill stage, resulting in prolonged time-to-first-token (TTFT). To address this issue, this paper proposes a method to reduce TTFT by leveraging a disk-based key-value (KV) cache to lessen the computational burden during the prefill stage. We also introduce a disk-based shared KV cache management system, called Shared RAG-DCache, for multi-instance LLM RAG service environments. This system, together with an optimal system configuration, improves both throughput and latency under given resource constraints. Shared RAG-DCache exploits the locality of documents related to user queries in RAG, as well as the queueing delay in LLM inference services. It proactively generates and stores disk KV caches for query-related documents and shares them across multiple LLM instances to enhance inference performance. In experiments on a single host equipped with 2 GPUs and 1 CPU, Shared RAG-DCache achieved a 15~71% increase in throughput and up to a 12~65% reduction in latency, depending on the resource configuration.

Large language models have been shown to perform well on a variety of natural language processing problems. However, as the model size and the input sequence's length increase, the rapid increase of KV Cache significantly slows down inference speed. Therefore GQA model, as an alternative to MHA model, has been widely introduced into LLMs. In this work, we propose a low-cost method for pruning MHA models into GQA models with any compression ratio of key-value heads. Our method is based on $\mathit{L_0}$ masks to gradually remove redundant parameters. In addition, we apply orthogonal transformations to attention heads without changing the model to increase similarity between attention heads before pruning training, in order to further improve performance of the model. Our method can be compatible with rotary position embedding (RoPE), which means the model after training can be fully adapted to the mainstream standard GQA framework. Experiments demonstrate that our strategy can compress up to 87.5% of key-value heads of the LLaMA2-7B model without too much performance degradation, just achieved through supervised fine-tuning.

Jailbreak attacks circumvent LLMs' built-in safeguards by concealing harmful queries within jailbreak prompts. While existing defenses primarily focus on mitigating the effects of jailbreak prompts, they often prove inadequate as jailbreak prompts can take arbitrary, adaptive forms. This paper presents RobustKV, a novel defense that adopts a fundamentally different approach by selectively removing critical tokens of harmful queries from key-value (KV) caches. Intuitively, for a jailbreak prompt to be effective, its tokens must achieve sufficient'importance' (as measured by attention scores), which inevitably lowers the importance of tokens in the concealed harmful query. Thus, by strategically evicting the KVs of the lowest-ranked tokens, RobustKV diminishes the presence of the harmful query in the KV cache, thus preventing the LLM from generating malicious responses. Extensive evaluation using benchmark datasets and models demonstrates that RobustKV effectively counters state-of-the-art jailbreak attacks while maintaining the LLM's general performance on benign queries. Moreover, RobustKV creates an intriguing evasiveness dilemma for adversaries, forcing them to balance between evading RobustKV and bypassing the LLM's built-in safeguards. This trade-off contributes to RobustKV's robustness against adaptive attacks. Large language models (LLMs) have gained surging popularity due to their unprecedented performance across various tasks. However, recent studies reveal that LLMs are vulnerable to a range of malicious manipulations, including training data leakage (Carlini et al., 2021), toxic content generation (Deshpande et al., 2023), and malicious fine-tuning (Qi et al., 2024). Of particular concern are jailbreak attacks, which represent a major threat to LLM security (Liu et al., 2023a).

As the use of large language models (LLMs) expands rapidly, so does the range of knowledge needed to supplement various LLM queries. Thus, enabling flexible and efficient injection of new knowledge in LLM inference is critical. Three high-level options exist: (i) embedding the knowledge in LLM's weights (i.e., fine-tuning), (ii) including the knowledge as a part of LLM's text input (i.e., in-context learning), or (iii) injecting the KV caches of the new knowledge to LLM during prefill. This paper argues that, although fine-tuning and in-context learning are popular, using KV caches as the medium of knowledge could simultaneously enable more modular management of knowledge injection and more efficient LLM serving with low cost and fast response. To realize these benefits, we envision a Knowledge Delivery Network (KDN), a new system component in LLM services that dynamically optimizes the storage, transfer, and composition of KV cache across LLM engines and other compute and storage resources. We believe that, just like content delivery networks (CDNs), such as Akamai, enabled the success of the Internet ecosystem through their efficient data delivery, KDNs will be critical to the success of LLM applications through their efficient knowledge delivery. We have open-sourced a KDN prototype at https://github.com/LMCache/LMCache.

The advent of pre-trained large language models (LLMs) has revolutionized various natural language processing tasks. These models predominantly employ an auto-regressive decoding mechanism that utilizes Key-Value (KV) caches to eliminate redundant calculations for previous tokens. Nevertheless, as context lengths and batch sizes increase, the linear expansion in memory footprint of KV caches becomes a key bottleneck of LLM deployment, which decreases generation speeds significantly. To mitigate this issue, previous techniques like multi-query attention (MQA) and grouped-query attention (GQA) have been developed, in order to reduce KV heads to accelerate inference with comparable accuracy to multi-head attention (MHA). Despite their effectiveness, existing strategies for compressing MHA often overlook the intrinsic properties of the KV caches. In this work, we explore the low-rank characteristics of the KV caches and propose a novel approach for compressing KV heads. In particular, we carefully optimize the MHA-to-GQA transformation to minimize compression error, and to remain compatible with rotary position embeddings (RoPE), we also introduce specialized strategies for key caches with RoPE. We demonstrate that our method can compress half or even three-quarters of KV heads while maintaining performance comparable to the original LLMs, which presents a promising direction for more efficient LLM deployment in resource-constrained environments.