Cache

Key-Value (KV) cache reduces inference latency in large language models (LLMs). Traditional prefix-based reuse has low cache hit rates across inference requests because it requires exact token and position matches. To improve efficiency, recent system optimizations introduce position-independent KV reuse, allowing KV cache to be reused whenever identical text chunks appear, regardless of their position in the sequence. We show this design introduces a new threat, KV Cache Hijacking. Since KV caches are retrieved by token match but encode the context in which they were originally computed, the KV tied to a benign-looking token chunk may encode an attacker-controlled prefix. When later reused in a victim query, this contaminated KV silently hijacks the model's behavior, even if no attacker-controlled text appears in the input. We introduce HIJACKKV, the first attack framework that systematically exploits this vulnerability, demonstrating its severity and practicality. HIJACKKV optimizes an attacker-controlled prefix, so that the KV computed for a subsequent common benign text encodes the attacker's goal, while the text remains unchanged for future cache hits. HIJACKKV achieves an average 94% success rate in a single attempt, remains effective under realistic constraints including low hit rates (10%) and frequent recomputation (50%), persists over multi-turn interactions, and transfers across models in black-box settings. We further provide design insights for building secure KV reuse systems.

We derive four memory-optimal inference artifacts for transformer attention using the Mathematics of Arrays (MoA), each following directly from the forward-pass Denotational Normal Form (DNF) of with the query-row index fixed to the current decode step. The artifacts are: (1)~a single-query decode DNF in which the $ψ$-reduction eliminates the $K^\top$ buffer algebraically, achieving $(d_k + nd_k+ nd_v+ d_v)\times4\,{B}$ Dynamic Random Access Memory (DRAM) traffic result numerically verified to $\|{err}\|_\leq2\times10^{-7}$; (2)~a C/OpenACC Graphics Processing Unit (GPU) kernel with Operational Normal Form (ONF) stride arithmetic and hardware-coalesced memory access, verified to $\|\mathrm{err}\|_\infty=0$ (exact IEEE-754 floating-point arithmetic); (3)~a multi-step KV-cache with $O(d_k+d_v)$ per-step append via MoA concatenation $\#$; and (4)~Grouped-Query Attention (GQA) and Multi-Query Attention (MQA) derived via $ψ$-selection, achieving a proven $\frac {h_q} { h_{kv} }$ reduction in KV traffic. All programs are verified against PyTorch scaled_dot_product_attention.

Long-context inference is central to modern large language model (LLM) applications such as retrieval-augmented generation and multi-document reasoning. To mitigate the growing inference cost, recent work has explored key-value (KV) cache reuse to reduce redundant prefill computation. However, existing reuse methods primarily focus on computation savings and overlook a critical bottleneck in long-context LLM serving: the cost of storing and accessing large KV caches. While KV compression appears to be a natural complement, naively combining compression with non-prefix KV reuse often leads to severe accuracy degradation. In this work, we propose C$^2$KV, a unified framework for non-prefix KV reuse that jointly optimizes KV extraction and inference-time concatenation. C$^2$KV learns a composable and compressed KV cache manifold that is explicitly designed to be position-agnostic. Our approach introduces a lightweight sidecar Extractor with learnable compression tokens and a structured attention flow, enabling modular KV representations that can be flexibly reused and concatenated without modifying the frozen base model. We further employ a compression-concatenation co-training strategy to align extraction-time representations with their downstream reuse behavior. Extensive experiments across multiple long-context benchmarks and model families demonstrate that C$^2$KV significantly reduces KV cache storage and transfer costs, achieving up to 17$\times$ inference speedup under long contexts, while preserving generation quality.

Block-wise diffusion large language models (dLLMs) decode sequentially at the block level, enabling effective KV-cache reuse across blocks but making inter-block decoding strictly serial. Prior work has attempted to unlock inter-block parallelism through post-training methods, but achieves only modest speedups and often degrades accuracy. We observe that self-correcting dLLMs offer a training-free alternative: token-to-token (T2T) editing can repair tokens drafted with a slightly stale upstream context, so a downstream block requires only an informative draft rather than a finalized predecessor. This turns block finality from a hard dependency into a scheduling resource. We propose \textbf{\flowblock{}}, a training-free parallel decoding framework built on two mechanisms. (i) \emph{Gated Wavefront Decoding} admits blocks into a bounded wavefront only when a readiness gate is satisfied, jointly refines active blocks via T2T editing, and commits blocks in order under a windowed block-causal mask that preserves exact frozen-prefix KV caches reuse. (ii) \emph{Heterogeneous Wavefront Packing} assigns each request an independent wavefront while packing asynchronous windows into dense, shape-stable batched forwards. Across different benchmarks, \flowblock{} improves tokens per second (TPS) over LLaDA-2.1 and LLaDA-2.0, two serial block-wise dLLMs, by up to 2.95$\times$ and 4.01$\times$, while reducing latency by up to 53.6% and 77.1%, respectively. It also improves average accuracy by 1.3 points. Compared with D2F, a training-based inter-block-parallel baseline, \flowblock{} achieves higher accuracy and up to 16$\times$ higher batched serving throughput.

As large language models (LLMs) process increasingly longer prompts, computation and KV-cache memory costs have emerged as major bottlenecks in inference systems. Existing input-level prompt compression methods address this, but rank each sentence by a scalar relevance score, treating the document as an unstructured pool of words and sentences. Under tight budgets, this causes theme collapse, where the dominant theme(s) of a document consumes the budget, discarding less-frequent yet task-relevant themes. Preserving thematic coverage instead requires allocating the budget across recurring themes rather than scoring sentences in isolation. To this end, we propose SALT, a model-agnostic extractive framework that organizes per-sentence keywords into a trie ordered by sentence frequency (SF), a lightweight, reusable proxy for document thematic structure. This trie-based organization smooths memory allocation and prevents dominant themes from monopolizing the budget. Multi-anchor retrieval activates trie nodes labeled by query keywords at any depth, and the trie persists across dialogue turns, supporting multi-turn use without re-encoding the document. By preserving document themes, SALT reduces the prefill computation and memory cost of long-context prompts while remaining composable with KV-cache methods that target decoding-time latency and memory.

We study whether \sigreg -- LeJEPA's anti-collapse objective -- can reshape representations during standard autoregressive language-model pretraining, and when the resulting geometry helps \kv-cache quantization. We train 110M-parameter models on 10B FineWeb tokens and report three findings. \textbf{(1)} At $λ{=}0.01$, \sigreg reduces hidden-state pairwise-cosine anisotropy by $38\%$ across three paired seeds. Perplexity increases by less than $0.35\%$ in every pair, with no consistent zero-shot loss. \textbf{(2)} This change does not propagate from hidden states to the \kv cache. Applying \sigreg directly to K and V during continued training, however, reduces mean cache anisotropy by $94\%$ across four checkpoints. A matched continuation without the \kv term leaves cache geometry nearly unchanged, and the frozen-trunk retrofits we tested do not reproduce the effect. \textbf{(3)} Under untransformed symmetric group-free quantization, direct \kv regularization is the only training condition that prefers per-channel scaling in all three seeds, and under that same 3-bit per-channel scheme the baseline incurs $4.3$--$7.9\times$ the directly regularized model's \dnll. Under the full simulated KIVI-style configuration (mixed arrangement, zero-points, grouped scales), however, all models reach near-parity, including when storage overhead is approximately matched. In this 110M regime, the training intervention helps when quantizer scales are coarse; the advantage vanishes under the tested combination of token-local grouping, mixed \kv scaling, and zero-points. To our knowledge this is the first training-time \emph{distributional} regularization of standard \kv-cache geometry evaluated against post-hoc cache quantization.

Production LLM deployments combine two cost-reduction primitives: prompt caching (a discounted rate for re-used token prefixes) and prompt compression (fewer tokens sent). The compression literature has standardized on query-aware methods that produce a different compressed prefix per query, mechanically invalidating the prefix-strict cache on every call. We characterize this cost empirically on Anthropic's Sonnet 4.6 API and find caching is far from the rho=1.0 ideal the literature assumes: Sonnet's cache has a two-tier architecture with a sharp threshold near 3,500 tokens, below which the hit rate plateaus at rho~0.83 across 30-call sessions. Our cost model predicts, and experiments confirm, that under realistic rho, query-aware compression beats naive caching at high compression ratios (r>=6). We propose Cache-Aware Prompt Compression (CAPC), pairing query-agnostic compression with explicit cache_control plus a tier-preserving ratio bound that prevents over-compression from pushing the cached prefix into the hot tier. CAPC is the cheapest strategy in 16/16 configurations on LongBench-v2, with mean savings of 49% over cache-only, 64% over query-aware compression, and 90% over vanilla, at quality within 0.05 of the uncompressed baseline. We validate CAPC on three production workloads: an enterprise tool-using assistant with a 94k-token schema prefix (51.7% cost reduction at r=3); a graphify knowledge-graph RAG pipeline across two codebases (9.3x vs cache-all on FastAPI, 2.4x on httpx); and the public tau-bench retail benchmark (50 tasks), where CAPC is the cheapest of four strategies with reward exactly equal to vanilla (both 36/50, p=1.00) while query-aware compression is the most expensive at +40.1% over vanilla -- the first production confirmation of the crossover model's negative-ROI prediction on a public benchmark.

We report a way to make a frozen small language model both more capable and dramatically cheaper at once, without changing any weights. Verified knowledge is deposited once as a byte-exact key-value (KV) state artifact and later restored, by graft, into a fresh inference context. The restore is bit-exact: under a pinned deterministic configuration, the grafted logits are byte-for-byte identical to a fresh computation (SHA-256 equality), with zero KL divergence and 100% argmax agreement over fifty samples. We show that own-position graft is the unique numerically exact operating point on a model with floating-point rotary encoding, and we verify byte-exactness on two model scales (12B, 31B) and two GPU targets, one through a pre-registered replay. On AIME 2025, a frozen Gemma-4-12B moves from 80.0% to 93.3% once a verified solution library is grafted, above its own 77.5% and its 31B sibling's 89.2% published anchors. On the recurring case, eight problems the base model never solves within a 401,026-token budget are answered from cached verified solutions in 61 total decode tokens, a factor of 6,574 fewer tokens and about 8,700x less energy; the capability claim proper rests on held-out transfer (7 of 7 at 31B). The same byte-exact store widens usable context from 32,768 to 2,854,766 tokens at zero extra accelerator memory, and moves byte-identical between machines of the same architecture. We describe the system at the behavior level; the engine is proprietary, and every reported number is backed by committed input and output hashes so the scoring can be re-checked without it.

We present cayleyR, an R package for solving permutation puzzles by detecting cycle intersections in Cayley graphs. The core algorithm performs an iterative bidirectional search: from both the initial and target permutation states, random operation sequences generate cycles in the Cayley graph of the symmetric group Sn; their intersection yields a connecting path. When no direct intersection is found, a distance-guided bridge selection narrows the gap, and the process repeats. The package targets the TopSpin(n,k) puzzle, whose state space is a Cayley graph of Sn generated by a cyclic shift and a prefix reversal. We describe the mathematical framework, the algorithm, and its implementation, which combines a C++ hash-indexed state store with optional Vulkan GPU acceleration. The software is publicly available on CRAN.

Attention-based KV cache eviction (H2O and its descendants) compresses the memory-constrained state of a long-context model by ranking tokens on accumulated attention mass, treated here as signal energy, and keeping the heaviest. On schema-dense input streams such as nested JSON, this score acts as a non-stationary filter that disproportionately retains noise: a non-content sink role (delimiters or whitespace) carries an order of magnitude more energy than any content role, and structural KEY tokens are over-retained at roughly 1.8x the rate of the answer-carrying VALUE tokens, collapsing exact-match accuracy from 88% to 0% at a 5% budget as the signal-to-noise ratio of the retained state degrades. A counterfactual experiment establishes that suppressing KEY tokens is the best deployable filter. Our retraining-free, role-conditional allocation over SnapKV's windowed score, governed by a single tuned hyperparameter, closes 63-98% of the H2O gap at sub-20% budgets and, at higher budgets, modestly matches or exceeds full-cache accuracy -- a small, seed-sensitive denoising effect (borderline significant at B=0.50; not distinguishable from zero at B=0.30 over four seeds). A 15 MB linear role probe supplies these labels at negligible inference cost, though matching parser-level downstream accuracy remains open.

The key-value (KV) cache has become the dominant memory cost of transformer inference. It grows with batch size, context length, and depth, and at long context it, rather than the model weights, sets the ceiling on throughput. Two families of methods reduce it. Low-rank methods factor two-dimensional slices of the cache, either per-head matrices or cross-layer feature blocks, and quantization methods lower the bit-width of every entry. Neither family exploits the fact that the cache at a layer is naturally a third-order tensor whose three axes, the heads, the tokens, and the features, carry very different amounts of redundancy. We take this tensor view directly. Our method, JoLT, applies a partial Tucker decomposition that compresses only the token and feature axes while leaving the head and layer axes intact, and then restores the energy that truncation discards with a Johnson-Lindenstrauss (JL) rotated low-bit residual. A single Lagrangian dual allocates the Tucker ranks and the residual bit-widths together, per layer group and separately for keys and values, under one byte budget. The result is a near-lossless 2-3x compression: perplexity, GSM8K accuracy, and RULER needle-in-a-haystack retrieval all stay at or within statistical noise of the uncompressed baseline on both a grouped-query-attention model (Mistral-7B-v0.3) and a multi-head-attention model (LLaMA-2-13B). At 2x, JoLT reconstructs the cache to relative Frobenius error 0.009 (K) and 0.006 (V) on both architectures, roughly an order of magnitude below cross-layer SVD and 4-bit quantization. A randomized-SVD variant, FlashJoLT, delivers a 5-13x compression-time speedup at matched quality.

Video diffusion models produce high-quality generations but remain slow at inference due to their sequential denoising procedure. Caching-based acceleration methods address this by reusing intermediate model outputs: leading dynamic approaches such as TeaCache, EasyCache, and DiCache accumulate a drift signal and skip expensive model evaluations when accumulated drift stays below a fixed threshold $τ$. This threshold controls an apparent tradeoff - raising it yields faster generation at the cost of visual quality, while lowering it preserves quality but sacrifices speed. We show this tradeoff is not fundamental; it is an artifact of holding $τ$ constant throughout denoising. We identify the existence of critical steps - timesteps where the drift signal changes rapidly - and show that applying a low threshold selectively at these steps while caching aggressively elsewhere recovers most of the quality of conservative caching at substantially higher inference speeds. Building on this insight, we propose ACID, a lightweight, training-free wrapper that monitors the rate of change of each method's existing drift signal to dynamically switch between a low and a high threshold. ACID is signal-agnostic and modular: it requires no retraining and plugs directly into existing dynamic caching methods without modifying their core mechanisms. Evaluated across three caching methods (TeaCache, EasyCache, DiCache) and three open-source video diffusion models (HunyuanVideo, Wan 2.1, CogVideoX), ACID consistently expands the Pareto frontier of visual quality versus inference speed beyond what any fixed threshold achieves. In particular, on TeaCache and HunyuanVideo, ACID achieves up to 2.16x speedup over the no-caching baseline, and up to 38% additional speedup over the conservative fixed-threshold baseline with negligible (<0.3 dB PSNR, <0.01 SSIM, <0.01 LPIPS) quality degradation.

We present a full-pipeline inference optimization for the MiMo-V2.5 model family, which combines Hybrid Sliding Window Attention (Hybrid SWA), sparse Mixture-of-Experts (MoE), and multimodal encoders. While Hybrid SWA can ideally reduce both attention compute and KVCache storage significantly compared to Full Attention, realizing these gains in production requires substantial engineering effort. We systematically optimize the KVCache system with layerwise prefetch, SWA-aware prefix cache trees, and specialized placement strategies, achieving strict $O(W)$ SWA storage and high cache hit rates. We further build GCache, a high-performance distributed cache infrastructure with RDMA-optimized networking, and develop a KVCache-affinity router to reduce computation while preserving load balancing. We also optimize for multimodal inputs, including GPU image preprocessing, parallel video decoding, and multimodal cache sharing. Together, these optimizations constitute the first large-scale LLM serving system in production that efficiently covers the Hybrid SWA + MoE + multimodal composite architecture.

On-device LLM inference faces a trilemma of response latency, limited hardware resources and user privacy. Full cloud inference delivers strong computing power but exposes user prompts and dialogue data, while standalone on-device inference is unfeasible for most consumer and embedded edge devices. This paper presents a privacy-centric edge-cloud collaborative LLM inference framework built on endpoint-authenticated KV cache. Local endpoints handle input preprocessing, embedding computation, adaptive feature optimization, KV cache authentication, speculative decoding and low-dimensional model head calculation, while the cloud conducts authenticated decoder inference, KV cache management, token verification and high-dimensional vocabulary projection. Endpoints fuse partial outputs, apply language-adaptive masking and sample target tokens. All transmitted data and truncated logits are quantized and AES-GCM encrypted for privacy, with core lightweight modules, draft parameters and cache access policies kept local to avoid leakage. The framework supports heterogeneous devices including CPU-only, GPU-equipped and embedded devices via optimized streaming, batching and quantized ONNX deployment. Evaluations demonstrate that the framework reduces per-token latency by up to 46.1% and downlink payloads by up to 67.4% over baseline split inference, retaining comparable performance to full cloud inference.

Large language model (LLM) agents accumulate heterogeneous context, including system instructions, plans, user turns, retrieved documents, tool outputs, and intermediate reasoning, whose key-value (KV) cache can become a major memory bottleneck. Existing eviction policies generally apply the same attention- or recency-based rule to every token, ignoring semantic structure already available to the agent orchestrator. We introduce MemDecay, a training-free, region-aware KV-cache eviction policy. MemDecay assigns tokens region-specific base priorities and decay rates, refreshes retention scores when tokens receive attention, and evicts the lowest-scoring pages under a fixed cache budget while allowing critical regions to be pinned. We also provide a procedure for calibrating decay rates from measured attention lifetimes. We evaluate MemDecay at approximately 450 and 1,700 token contexts using Qwen2.5-1.5B and 3B. Across all settings, attention lifetimes differ by an order of magnitude across regions: system-token half-lives range from 148 to 189 decoding steps, compared with 14 to 16 for scratchpad tokens. Pinning preserves system-region facts at full-cache accuracy in every setting, while no baseline preserves more than 13 of 24. Region-aware retention remains effective as context grows, whereas recency-based retention collapses. Accumulated-attention retention performs better on unpinned content, however, and ablations identify attention-score normalization as the main limitation of the current formulation. These results establish semantic prompt structure as a robust signal for KV-cache management while clarifying how it should be combined with attention-based importance.

Context engineering decides what information a model carries forward, and current designs meter it in tokens: compressing the past into a bounded recurrent state, keeping a key-value entry for every token, or imposing a fixed budget through a window or eviction rule. All three make the token the unit of memory even when the stream is redundant and the task depends on the distinct information it carries. Building on a companion mechanism paper that opens a cache slot only when an incoming key is novel, so memory scales with the number of distinct items rather than tokens, we develop that allocate-on-novelty cache as a working-memory component and organize context by how a task depends on the past: recall-carried information belongs in a content-addressed novelty cache, summary-carried information in a recurrent state, and locality-carried information in a recency window. The claim is empirical and bounded. On a matched character-level control, novelty-gated attention reaches full-attention performance while attending to about half the tokens, and coupling the cache with a state-space summary matches full-attention coupling at that reduced cost; the advantage grows as context lengthens, while a sliding window is preferable on short, locality-dominated spans. On next-code prediction over synthetic Medicare claims the coupled component leads full attention and every fixed-budget eviction policy at a thousand-event horizon, whereas cost forecasting over the same stream is summary-carried and the cache is neutral. The retained memory is an inspectable table of templates, codes, drugs, or places rather than an opaque state. The experiments are small-scale and use only public data; they establish the primitive that context can scale with distinct information rather than tokens, in a working memory that is content-addressable and auditable.

A persistent interactive world model keeps its running state resident on the GPU that serves it: a multi-gigabyte attention cache, almost all of it rewritten at every generation step. That state cannot be recomputed in interactive time or approximated without changing the world, so a live session pins its device. The pin is a scheduling problem. WorldMove moves a live session under one guarantee: the destination is bit-identical to the source, or nothing is installed. It relocates the cache in 18.8 ms same-node, 101x faster than save/load. It holds a checksum-verified 92.1-94.8 Gb/s on a 100 Gb fabric. At that rate the cache fits inside one interactive block. Migrating an actively generating session, it converges at a block boundary and the destination continues the world bit for bit. An admissibility condition decides each move. The move must complete inside the readout horizon, over bandwidth that covers the state plus its dirty rate. Lifted to a fleet schedulability test, it governed a consolidation loop that executed 48 of 48 migrations bit-identical across two providers. Two constraints are structural. Bit-exactness survives only inside a controlled configuration of one GPU architecture, so moving the state is the only way to preserve it exactly in interactive time. Verification cannot hide inside the wire on this fabric. Receive-path checksums stall the transport at protocol timescales under fan-in, and unscheduled incast silently collapses a receiver while every delivered byte stays correct. An incast-aware admission controller holds zero misses to 1.4x offered load and sheds overload as rejects. A lossless GPU codec widens the admission gate to fabrics raw motion cannot use. We exercise the serving loop and the mover separately, each end to end. Their composition on one fabric is unbuilt. Exact-state elasticity is a joint scheduling problem over transport and verification.

KV-cache compression methods are predominantly evaluated with the query appended to the context before compression -- a query-aware protocol. Yet the economic case for a compressed KV cache is reuse: compress a document once, answer many future questions against it. In that deployment, compression must happen query-agnostic -- before any question is seen. We present a matched-budget audit of six published compression methods against three trivial baselines on three open 7-9B models (144,300 paired evaluations on RULER-8192; 40,800 on LongBench; 50,000-resample paired bootstrap throughout). Everything is held fixed -- model, compression ratio, instances, decoding -- except the scoring rule. Three findings. (1) Query visibility changes the rankings: under the agnostic protocol, of the five audited methods that share a common attention backend, only KeyDiff beats a best-of-3 trivial baseline consistently (31 of 36 cells), and the most widely deployed method, SnapKV, loses to "keep the start and the recent window" on average (-0.066). (2) The per-method drop between the two protocols is ordered consistently with how visible the question is to each method's scoring signal, legible in its source code: from Delta=+0.198 for SnapKV (the question sits inside its 64-token observation window) down to Delta=+0.011 for KeyDiff (its score contains no query term at all).

Containerization has become increasingly essential in the machine learning (ML) domain, providing reproducibility, portability, and environment consistency. While prior studies have analyzed Dockerfile structures and best practices, none have examined ML projects in depth to reveal how the iterative nature of ML workflows influences container footprint, build performance, and caching behavior. We present the first large scale empirical study of 1,993 ML related Dockerfiles, combining quantitative analysis of container roles in ML projects and build dynamics with a qualitative investigation of refactoring practices. Results show that containers serve distinct roles across training, inference, and infrastructure. Containers are typically large, averaging 10.27 GB in size, and require long build times of about 8.84 minutes. We find that 44.4% of commits trigger rebuilds, primarily due to context file changes (96.4%), with experimentation being the main motive behind those commits that initiate rebuilds. Despite partial cache reuse, 71% of rebuild work is wasted on redundant computation. From stable projects, we identify 7 recurring ML-specific Dockerfile refactoring patterns that improve build efficiency and reduce container footprint.

Fixed-state sequence models compress an unbounded past into a bounded state, which caps their associative recall at roughly the state dimension; attention escapes the cap by keeping a key-value entry for every token, at quadratic compute and a cache that grows with the sequence. We study the middle ground: a sparse cache that allocates a slot only when an input is novel, so its size tracks the number of distinct items rather than the number of tokens. The allocation rule is the DP-means clustering rule, the small-variance limit of a Dirichlet-process mixture, used not as latent-variable inference but as the key-value memory operator for a deep recurrent backbone. We develop it in two forms, a static cache with a fixed concentration and a surprise-adaptive variant whose concentration follows the recent novelty rate. On a controlled associative-recall benchmark with redundancy we show that the cache matches full-attention recall while storing only the distinct items, that it dominates a fixed-budget eviction cache on the recall-versus-size frontier, and that on a state-space backbone it answers both a recall query and a long-range aggregate at the lowest memory of any model tested. The allocation is learnable end to end: a two-parameter novelty-threshold gate trained on the task loss alone recovers the rule exactly, whereas an over-parameterized gate fails, so the operative ingredient is the inductive bias rather than capacity. The evidence is a family of controlled mechanism studies at modest scale, with the distinct-items property confirmed on four real streams (recommendation, systems logs, clinical events, and insurance claims); a real-backbone, real-corpus language validation is pursued in a companion study.

Block sparse attention is a hardware friendly way to alleviate the key-value (KV) cache read bottleneck in large language models (LLMs). However, it is not prevalent among leading open-weight LLMs, which rely instead on dense attention or fine-grained selection, thereby motivating our analysis. We study DeepSeek's Native Sparse Attention (NSA) as a representative method, whose three-branch design lets us isolate block selection, the most challenging and consequential stage. We formalize selection and reduce it to ranking blocks by a single quantity, the attention mass: the sum of a block's attention scores. We show that if selection retrieves the blocks with the largest attention mass, block sparse attention can match the quality of dense attention. However, computing the exact attention mass requires reading every key, so the problem of block selection ultimately reduces to approximating this mass from a compact summary instead of the full keys. Via a cumulant expansion, we show why existing methods falter: their selection strategies attempt to estimate the attention mass, but are confined to a first-order approximation. Therefore, we propose COBS (Cumulant Order Block Sparse Attention), an attention method that builds on NSA, incorporating a novel selector that stores a compressed second-order statistic per block. On the 32k RULER long-context retrieval benchmark, COBS raises the NSA baseline's mean score from 0.2999 to 0.8195, approaching dense attention at 0.9040 and closing about 86% of the gap, while using only 1.21x the KV cache read traffic of the NSA baseline and 15.15x less read traffic than dense. The same model preserves short-context behavior and attains lower position-wise negative log-likelihood (NLL) than dense attention in our comparison.

Despite the rapid advancements of large language models (LLMs), LLM serving systems remain memory-intensive and costly. The key-value (KV) cache, which stores KV tensors during autoregressive decoding, is crucial for enabling low-latency, high-throughput LLM inference serving. In this survey, we focus on system-aware KV infrastructure for serving LLMs (abbreviated as sKis). We revisit recent work from a system behavior perspective, organizing existing efforts into three dimensions: execution and scheduling (temporal), placement and migration (spatial), and representation and retention (structural). Furthermore, we analyze cross-behavior co-design affinity and behavior-objective links, highlighting future opportunities. Our work systematizes a rapidly evolving area, providing a foundation for understanding and innovating KV cache designs in modern LLM serving infrastructure.

Large language models, and the agents built on them, spend an ever-growing share of their compute and memory on remembering: caching attention keys and values, carrying long prompts, maintaining recurrent state, and storing what happened in previous turns and sessions. Because none of this memory is free, four largely separate research communities have each learned to compact it. They evict or quantize the KV cache, prune or distill prompts, bound architectural state, and consolidate agent memory. We argue that these are instances of one problem: a rate--distortion decision about what context-derived information to retain versus discard, at what fidelity, under a resource budget, so as to preserve downstream task utility. We make this lens precise with a single compaction objective and a layer-agnostic lower bound, use it to build a seven-axis taxonomy that classifies methods from across the stack uniformly, and use it to transfer mechanisms between layers that have never been connected, from serving-stack KV management to agent long-term memory. Two patterns hold across the survey. At every layer the signal that decides what to keep is attention magnitude or recency, and it fails in the same way everywhere, by discarding, before the query is known and with no way to undo it, information the query later needs. And while compression is measured carefully on single-turn long context, the repeated compaction that agents actually perform is almost never measured, and no benchmark holds one budget axis across all the layers at once. We turn both observations into a benchmark proposal, a small reference experiment, and a set of compaction-aware design principles, and we map the open problems.

The key-value (KV) cache dominates the memory cost of long-context autoregressive inference, and a growing body of work compresses it through quantization, eviction, or offloading. We study a complementary question: once a position's KV state has been quantized to codebook indices, how should the resulting symbol stream be stored, and can the storage layer do more than store? A family of contractive iterated-map codes that serialize a symbol sequence into a sequence of low-dimensional real vectors is revisited, and it is shown that they form a natural archive format for a quantized KV cache with the following features. The method provides exactly the access pattern a growing cache requires. It is lossless, it runs in linear time, and supports O(1) random access and O(1) amortized append. A controlled study of the quantizer feeding this archive is conducted on GPT-2 with 1024-token contexts. Keeping a small exact window (4 attention sinks + 32 recent tokens) and archiving the rest, per-head residual vector quantization reduces the archived cache by 36-54x relative to an fp16 cache at a perplexity cost of 11-15%, and we quantify a sharp key/value asymmetry -- quantizing keys is roughly 4x more damaging than quantizing values, consistent with prior low-bit KV work -- and use it to allocate bits in a hybrid scheme. Finally, we show the archive is simultaneously a search index: approximate substring queries execute directly on the stored vectors, and matched context is decoded from the matched vector without ever materializing the surrounding text. We release all code; every number reproduces from a single command on a laptop CPU.

Speech large language models (Speech LLMs) typically encode speech into sequences far longer than text, creating a major efficiency bottleneck during autoregressive decoding. A common remedy is to compress the speech sequence at the adapter level to remove temporal redundancy before it enters the LLM; however, such early downsampling risks discarding fine-grained information that cannot be recovered. We propose SpeechKV, which applies a learned pooling to the KV cache of speech tokens inside the LLM. This design allows the LLM to fuse speech and text internally while directly accelerating decoding. Trained on 71K hours of speech data, SpeechKV compresses the speech to approximately text-level granularity yet maintains performance on par with or even slightly better than the uncompressed baseline, with relative gains of 6.6% on out-of-domain entity recognition and 2.3% on OpenASR, while delivering at least 1.49 times decoding speedup that scales with audio length.

Long-context language model inference is increasingly limited by the memory bandwidth and capacity required to store key-value caches, yet existing compression methods often apply uniform budgets across layers or tokens and degrade retrieval when lexical cues and semantic states require different preservation. We introduce DepthWeave-KV, a token-adaptive cache compression method that factorizes key and value states across neighboring transformer layers using shared low-rank channel bases while retaining lightweight token-specific residuals where attention behavior is sensitive. DepthWeave-KV combines cross-depth residual factorization with a token-conditional depth router that allocates higher reconstruction rank to instruction-bearing and retrieval-critical tokens, and uses calibration-free online error tracking from attention-output probes to adapt compression during generation without retraining the base model. A fused CUDA implementation jointly performs basis lookup, residual dequantization, and attention projection to reduce decode-time memory traffic. Across LongBench, Needle-in-a-Haystack, L-Eval, and long-form QA and summarization benchmarks, DepthWeave-KV achieves near-full-cache task quality with substantially lower memory use, improving average score and retrieval accuracy over prior compressed caches while reaching 8.3x KV memory reduction and 72.8 tokens per second at 64K context.

Long-context LLM inference is increasingly limited by the memory and bandwidth cost of KV caches, yet aggressive compression can remove the layer-specific evidence needed for retrieval and multi-step reasoning. We introduce FreqDepthKV, an inference-time cache compression method that factorizes adjacent-layer KV states into shared low-frequency depth components and sparse high-frequency residuals. A lightweight online probe assigns attention heads to shared-depth, residual-depth, or exact cache modes according to their contribution to reconstruction-sensitive attention logits, allowing the compression policy to adapt to prompt structure without retraining. Across long-context question answering, needle retrieval, summarization, and code generation benchmarks, FreqDepthKV preserves task accuracy under substantially smaller cache budgets. With a 32k-token prefill window, FreqDepthKV reaches 58.3 Exact Match, 63.0 F1, 32.5 ROUGE-L, and 48.1 pass@1, closely matching full KV while outperforming prior compressed-cache methods. It also improves decoding throughput to 70.4 tokens/s, reduces TTFT to 2.06 seconds, and lowers peak KV memory to 6.2 GB, achieving a 3.9x effective compression ratio.

Vision-Language-Action (VLA) models have emerged as a promising approach for generalizable robotic manipulations. In particular, flow matching-based VLA models have shown remarkable success due to their capability to generate precise and smooth action sequences and capture multimodal distributions. However, the iterative denoising process in the action head acts as a major computational bottleneck, posing a critical challenge for real-time deployment. To address this challenge, we propose ActionCache, a plug-and-play external cache that opportunistically reuses past intermediate actions to warm-start generations from the vicinity of target actions, thereby drastically reducing the inference latency. Specifically, ActionCache stores the intermediate actions with compact multimodal keys, which enables retrieval from similar past contexts across different episodes or even different tasks. Experimental results in simulation and real-world environments demonstrate that ActionCache maintains high task success rates in a low-latency regime, achieving inference acceleration of up to $11.75\times$ and $34.43\times$ for representative flow-based VLA models, $π_{0.5}$ and GR00T-N1.6, respectively.

Conditional computation can decouple language model quality from per-token inference cost, yet leading techniques act on a single axis in isolation: Mixture-of-Experts (MoE) sparsifies the FFN, Mixture-of-Depths (MoD) skips whole transformer blocks, and KV-cache quantization compresses attention memory. We argue these three decisions (attention resolution, expert selection, and cache bit-width) are strongly coupled and should be made jointly: a token rare enough to warrant full attention may also need high-precision caching regardless of which expert processes it. We introduce TriRoute, a single lightweight controller shared across all three axes that, for every token at every layer, emits a coordinated policy: (i) an attention mode (skip/local/full), (ii) a sparse set of FFN experts (with a null expert recovering MoD), and (iii) a KV-cache bit-width. The controller trains end-to-end via a heterogeneous relaxation (Gumbel-Softmax with straight-through estimation for categorical decisions and load-balanced top-k gating for experts) under a Lagrangian budget constraint that turns the average compute and memory cost into a controllable knob. We identify a cross-axis routing-collapse cascade in naive joint training, where collapse on one axis propagates to the others, and address it with per-axis normalization and a coupling-aware balancing loss. On decoder-only models from 160M to 1.3B parameters at compute-optimal token counts, TriRoute Pareto-dominates the best independent MoD+MoE+KV-quantization combination at matched inference FLOPs and memory, while better preserving tail-case robustness on rare entities, code, and arithmetic that pure perplexity optimization erodes. Post-hoc analysis reveals interpretable structure: the controller allocates full attention and high-precision cache to sentence-initial positions, rare subwords, and named entities, while cheaply routing function words.

Key-value (KV) cache growth is a major bottleneck in autoregressive decoding, as memory and bandwidth scale linearly with context length. Existing KV eviction methods often rely on static heuristics or proxy scores, which poorly track future token utility and cause brittle eviction as relevance shifts. To address this, we introduce KVpop, which learns a fixed-budget KV eviction policy by directly supervising the keep-or-drop decision. The scorer is trained against a novel future-attention target, computed efficiently without materializing dense attention maps. We further introduce a delayed memory-based scorer that, uniquely among learned eviction methods, defers scoring for a fixed number of steps to exploit near-future context. On AIME and HMMT mathematical reasoning, KVpop retains 98% of full-attention performance on Qwen3-4B at 75% KV cache compression and 97% at 88% compression, consistently outperforming established eviction baselines. Qwen3-8B shows even stronger results, reaching near-full teacher performance. These results show that supervising eviction with future-attention signals cuts memory costs while maintaining quality.

Semantic caching reduces the latency and cost of retrieval-augmented generation (RAG) by serving cached answers to semantically similar queries, but most existing methods do not model the time-varying freshness of open-web evidence. We present FreshCache, a three-tier semantic cache that treats cache reuse as a risk-constrained temporal inference problem: before approving a cache hit, FreshCache estimates the probability that the cached result is stale using a fitted exponential decay model enhanced by a learned MLP, and approves reuse only when that probability falls below a per-tier error budget across answers (epsilon = 0.10), URL lists (epsilon = 0.20), and page content (epsilon = 0.35). This allows the system to degrade gracefully as entries age rather than forcing a binary choice between a stale hit and a full pipeline execution. We introduce FreshCache-Bench, a benchmark of 8,072 base queries across five freshness classes with ground truth staleness labels drawn from real web snapshots at 1, 12, 24 hours, and 7 days after a baseline crawl, expanded to 31,201 queries via paraphrase generation. At the 24-hour evaluation window, FreshCache_MLP achieves 97% search API savings at 0.1% hash-based stale error, and an LLM-judge evaluation on 396 confirmed change pairs shows that only 34.3% of detected content changes actually affect answer correctness, placing true answer-affecting stale error at approximately 0.034%. The rule-based FreshCache achieves 98% search savings at 3.3% stale error under a temporal holdout calibration, outperforming SemanticTTL (14.9% stale, 72% saved), vCache (7.2% stale, 47% saved), and SCALM (5.2% stale, 96% saved). Ablations show the temporal risk gate accounts for an 11.6 point reduction in stale error over similarity-only reuse, and the learned MLP reduces stale error a further 3.2 points over the rule-based model.

Diffusion language models (dLLMs) generate text by iteratively denoising a masked response and can commit multiple output positions per model invocation. Their bidirectional attention prevents exact autoregressive-style KV caching, since committing one position shifts the KV activations of all others. Approximate caching techniques such as Fast-dLLM and dKV-Cache refresh KV activations repeatedly and reuse them across intervening decodes, inducing a repeated prefill/decode structure. This makes AR serving mechanisms relevant to dLLMs, but not directly applicable. dLLM decodes are block-sized rather than token-sized, prefills recur, and bidirectional attention precludes the chunked prefill mechanism used for stall-free colocated serving. We present Sangam, a serving system for cached dLLM inference. Sangam introduces a deficit token-budget scheduler that admits in-flight decodes first, admits whole indivisible prefills only when the accumulated token budget allows, and carries unused budget forward. This achieves amortized stall-free scheduling. Disaggregated serving avoids prefill-decode interference but suffers from prefill/decode resource partitioning problem. Sangam adopts a hybrid serving strategy, overflowing prefills onto decode workers to relieve prefill under-provisioning, and uses the same deficit-budget scheduler to protect those workers' decodes from the overflow. We show that like AR serving, dLLM serving design space is governed by prefill-decode interference and prefill/decode partitioning. Colocated serving is most effective on decode-heavy workloads, cutting mean latency by 9-20% over hybrid execution on LLaDA-8B ShareGPT; while hybrid execution is most effective on prefill-heavy workloads, cutting mean latency by 8-20% over colocated execution on Dream-7B arXiv. Sangam is available at https://github.com/UT-InfraAI/sangam.

With the rise of small quantized GGUF-based language models and their increasing use for on-device inference tasks, we have seen the growing need for an approach capable of reliably delivering these models at scale even under severe memory bandwidth constraints such as those imposed by pure CPU implementations. Fixed-depth speculative decoding has emerged as one promising technique, but in practice, it often leads to performance degradation due to either bandwidth saturation, instability, or even catastrophic resource exhaustion resulting in system failure. To overcome this problem, we introduce AdaptiveSD, a fully runtime-adaptive speculative decoding framework aimed at ensuring robust, reliable execution across the spectrum of model types and workloads. Our solution consists of four tightly-coupled components working together in a continuous feedback loop: a Runtime Monitoring Engine tracking multiple signals relevant to ongoing computation, an Adaptive Draft Controller enforcing an eleven rule policy hierarchy prioritizing system resource preservation over raw draft count, a Dynamic Policy Engine employing a suite of heuristic and reinforcement learning techniques to dynamically modify policies depending upon workload behavior, and finally, a KV Cache Coordination Layer managing cache states with fine-grained control through INT8 shadow buffers and position aware evictions. While conventional approaches focus solely on maximizing throughput, we instead assess the effectiveness of our approach based on several key metrics including wasted drafted compute and inter-token latency dispersion alongside standard measures of speculative efficiency.

Real-time interaction models -- Moshi, MiniCPM-o, Qwen-Omni -- turn serving into a periodic real-time task: on every frame a session ingests streaming audio and must respond by a recurring wall-clock deadline, while its KV cache grows monotonically and stays pinned for the whole conversation. This regime hides a dangerous failure mode. On a real full-duplex stack, sustained load does not degrade serving gracefully: it falls off a cliff, jumping in one step from milliseconds per frame to a stalled engine when accumulated session state exhausts the KV pool. The collapse is metastable -- identical five-minute runs collapse or survive on run-to-run variance -- and silent: latency and deadline-miss metrics read healthy throughout. We show one move restores both stability and observability: bound each session's resident state, and latency starts telling the truth. Metronome's in-engine KV window eliminates the collapse (0/20 vs. 14/20 runs across two batches) and turns per-frame latency into a monotone load signal, on which an online admission controller discovers the schedulable concurrency; without the window, the identical controller over-admits into the wall. A first-order model predicts the collapse time within a few percent on the headline model, and a quality probe validates the bound's design by ablation: the window alone is quality-free in turn-based decoding, and its few pinned attention-sink tokens are what keep free-running generation healthy. Everything is measured end-to-end on real audio, across four interaction models on one GPU.

Long-context inference is increasingly common in large language model (LLM) serving, driven by retrieval-augmented generation and agentic systems. In disaggregated inference, these workloads require transferring large Key-Value (KV) caches across the network, where decoding cannot begin until the transfer completes. Recent KV quantization techniques reduce data volume and alleviate this bottleneck, but existing schemes fail to achieve both low network-exposed latency and high inference accuracy. We challenge the assumption that the KV cache is an indivisible unit that must be fully received before use. We leverage the observation that different bits in the KV cache contribute unequally to attention computation and inference precision: the most significant bits capture the coarse structure of attention and the least significant bits refine precision. This property enables partial use of the KV cache during decoding. We present Lynx, a system that enables progressive, split-stream KV transfer by partitioning the KV cache into a high-priority Anchor stream carrying the most significant bits and a low-priority Residual stream carrying remaining precision. Decoding begins upon receipt of the Anchor stream and proceeds speculatively while the Residual stream is transferred concurrently, followed by verification that ensures equivalence to higher-precision decoding. Across multiple models and serving workloads, Lynx achieves Time-to-First-Token (TTFT) comparable to aggressive 4-bit KV quantization, while matching the accuracy of high-precision (BF16) inference, improving TTFT over standard 8-bit KV quantization by up to $1.43\times$ and improving accuracy over state-of-the-art by up to $5.1\%$.

Transformer inference on long sequences is expensive because softmax attention repeatedly reads from a large KV cache. The prevalent approach to this bottleneck is KV cache compression, which replaces the full cache with a compact summary. Despite its practical importance, the design of such summaries is largely driven by empirical experimentation. On the theoretical side, existing results show that KV cache compression can be impossible in the worst case, but offer little systematic guidance for designing algorithms in regimes where accurate compression is possible. We bridge this gap by characterizing the minimax risk of KV cache compression in terms of the intrinsic compressibility of a cache, revealing when and how accurate compression is possible. These results yield novel design principles for KV cache compression under causal masking that map efficiently to prefill and autoregressive decoding while achieving minimax-optimal risk. We instantiate these principles in a practical algorithm and report promising performance on LongBench in targeted experiments. Overall, our results provide a principled avenue for practical KV cache compression with theoretical guarantees.

TriAttention is a recent method for shrinking the KV cache of long-reasoning LLMs: it scores each cached key by how much attention it is likely to receive and evicts the lowest-scoring ones. Because a key does not know how far away its future queries will sit, the score is averaged over a ladder of 17 possible future distances. We point out that this average is free: the future distance enters the score only through the position-dependent rotation, so the whole 17-fold average collapses--exactly, by a one-line algebraic identity--into a single per-band weight that is computed once, offline. Scoring a key then costs one evaluation instead of seventeen, with no change to which keys get pruned. The saving is modest and lives entirely in TriAttention's pruning-score computation, not in the attention kernel; we present it as a small, exact complement to their method, and we confirm the identity numerically.

Multi-agent latent reasoning composes agents' KV-caches into one context for a final agent. Prior work (Agent Primitives) does this by concatenating caches along the sequence axis with RoPE re-encoding, which we call BagMerge. BagMerge is non-commutative, and the best input ordering is unpredictable, shifting with the regime, the latent-step budget, and the model scale. We make this exchange a convergent replicated state. First, CanonicalMerge fixes the layout by content: ordering caches by mean K-norm at a middle layer renders the merged cache byte-identical under any input permutation, verified algorithmically (arity N<=5) and bit-for-bit on real Qwen3-1.7B and 4B state. Second, we separate the replicated state from decode-time layout: the state is a set of content-addressed latent fragments whose merge is set union, a state-based CvRDT (commutative, associative, idempotent, absorbing), and CanonicalMerge is its deterministic render. Because the render is byte-equivalent, every N=2 accuracy number carries over unchanged and re-delivered duplicates are absorbed rather than re-concatenated. On a partitioned-reasoning benchmark, CanonicalMerge matches the best BagMerge ordering in every regime-by-budget-by-ordering cell without knowing which order is best, trading a small, statistically insignificant accuracy margin for an unconditional structural guarantee. The behaviour transfers to real multi-document QA (HotpotQA), while the closest training-free output-fusion baseline (PackLLM) loses by 45 points at matched budget, placing cache-level merging in a regime distinct from output-level fusion. Finally, at k>2 the approach transports and colocates latent traces but does not by itself compose them, which we characterize to motivate future work.

The deployment of Large Language Models (LLMs) with extended context windows is increasingly constrained by the linear growth of Key-Value (KV) cache memory. Vector Quantization (VQ), particularly Residual Quantization (RQ), is a promising approach for pushing KV cache storage toward the sub-1-bit regime by progressively encoding residuals with small codebooks. However, most VQ methods still rely on standard $\ell_2$ $K$-means as the core codebook-learning primitive. We identify a subtle high-dimensional issue of this primitive: Euclidean centroid averaging can induce centroid shrinkage, which weakens the angular alignment term in the $\ell_2$ distortion and makes directional preservation harder. To address this issue, we propose Gain-Shape $K$-means (GSKM), a drop-in replacement for $K$-means that improves directional fidelity while matching, and in some regimes improving, $\ell_2$ distortion. We then build Gain-Shape Residual Quantization (GSRQ) by incorporating a weighted extension of GSKM into an RQ pipeline. On LLaMA-3-8B, GSRQ substantially improves over strong KV cache quantization baselines across bit rates. At 1-bit, it improves the average accuracy across LongBench tasks from 11.34 to 33.54, a gain of 22.20 percentage points over VQLLM.

Video reasoning requires Large Multimodal Models (LMMs) to remain grounded in dense evidence, yet existing systems largely adopt "read-once, generate-many" paradigm, in which visual grounding weakens during generation. This phenomenon has been widely observed and is known as Visual Anchoring Decay. To fill this gap, we introduce Latent Video Cache (Latent-VC), a recurrent latent visual cache inserted into the decoder to preserve compact visual memories throughout reasoning. The cache is trained with supervised contrastive cache alignment and vision-grounded GRPO with a latent grounding reward, while maintaining strict train-inference alignment through native decoder hidden states. Built on Qwen3.5-9B, Latent-VC consistently outperforms strong CoT and SFT+GRPO baselines across six video benchmarks, with especially clear gains on grounding-intensive and long-video tasks. In addition, it also achieves higher accuracy with substantially shorter responses, suggesting that latent visual caching improves video reasoning by preserving visual evidence rather than relying on longer textual chains.

Long-context LLM services now sustain prompts with hundreds of thousands to millions of tokens, making the key-value (KV) cache a first-order serving cost. Because the cache grows linearly with context length, it can exhaust GPU memory, force smaller batches, and reduce serving throughput. Prior KV cache compression techniques typically target only the sequence dimension or only the channel dimension, which leaves limited headroom as context windows scale. Compressing both dimensions promises higher memory reduction, but applying the two forms of compression directly leads to significant accuracy loss. This paper introduces MosaicKV, a dynamic two-D (dimensional) KV cache compression system for extremely long-context serving. MosaicKV uses dynamic two-D compression to address the accuracy challenge, exploiting the non-uniform importance distribution of elements within the KV cache. Instead of applying one compression pattern globally, MosaicKV identifies important elements for each KV vector and selects compression strategies at the granularity of KV cache segments. To address the performance challenge, where fine-grained sparsity and compression management overhead can offset the gains from compression, MosaicKV introduces compressed KV cache management. This mechanism uses underutilized GPU and CPU resources to maintain compressed KV caches and accelerate attention computation. Evaluation on an H800 GPU with multiple LLMs shows that MosaicKV delivers up to 16x attention speedup, 4.8x lower decode latency, and 7.3x higher throughput than the uncompressed baseline. At the same time, it reduces memory usage by 3x and incurs only 1.76% average accuracy loss on LongBench and RULER.

Autoregressive (AR) streaming models have emerged as a powerful paradigm for long video generation. However, the linearly growing Key-Value (KV) cache poses a significant bottleneck, leading to memory overload and degraded inference throughput. A common compression method is to drop redundant KV tokens, which often breaks long-range dependencies, resulting in temporal flickering and identity loss. In this paper, we propose Instance-Specific Parametric Absorption (ISPA), a novel framework that shifts the KV cache compression from discarding to distilling. The core idea is to transit a subset of layers from Full-Attention (F-Layers) to memory-efficient Local-Attention (L-Layers) by "absorbing" historical context into the model's weights. Specifically, during a brief warmup phase, ISPA monitors the output discrepancy between global and local attention. At the transition point, we solve a closed-form least-squares problem to compute an instance-specific weight modulation that compensates for the missing history. Experiments across architectures (1.3B to 14B) demonstrate that ISPA can remove up to 50% of the KV cache with near-lossless visual quality. We hope this perspective encourages future work to explore parametric memory consolidation beyond external token-level cache management for streaming generative models.

LLM agents increasingly rely on retrieval buffers to store and reuse past experience, yet the cache management policies governing these buffers remain largely ad-hoc. We formalize this as an online semantic cache replacement problem with switching costs, where items are matched by embedding similarity and hit quality is continuous rather than binary. Through experiments on two datasets from MemoryBench-Full (LoCoMo, DialSim) with 8 replacement policies, we reveal a surprising finding: classic heuristics (LRU, LFU) \emph{consistently underperform} the naive FIFO baseline on semantic workloads, due to the absence of temporal locality and frequency concentration. We propose SOLAR, a learning-augmented framework that derives modification timing from regret accumulation (achieving $\sim$17% modification rate) and content selection from Bayesian online learning over implicit retrieval feedback. We prove SOLAR achieves a constant competitive ratio $\leq 3$, independent of cache size and horizon (vs.\ $Ω(K)$ for FIFO), and eviction regret $O(\sqrt{KT\log T})$, matching the $Ω(\sqrt{KT})$ lower bound up to logarithmic factors. Experiments demonstrate 5--75% relative improvement over FIFO at tight cache sizes, with a clearly characterized phase transition at the working set boundary. Synthetic experiments with 5000-item pools further reveal an inverted-U relationship between pool size and retrieval quality, justifying capacity constraints as a retrieval noise phenomenon rather than a storage limitation.

The key-value (KV) cache has become a first-order memory object in LLM serving rather than a temporary per-request tensor. This survey classifies more than thirty KV-management systems and frameworks using four axes: locality, lifetime, ownership, and substrate. The axes reveal five architectural archetypes -- local-paged, disaggregated-pipeline, shared-store, memory-pool, and hybrid-tier. Once workload and hardware are fixed, ownership accounts for much of the remaining design variance among distributed systems. The survey also audits current evaluations and identifies seven missing KV-specific measurements, linking them to open problems in fault tolerance, isolation, tiered eviction, speculative decoding, MoE serving, and shared-cache semantics.

Long-context Large Language Model inference is severely bottlenecked by the massive Key-Value (KV) cache, yet existing sparse attention methods often suffer from static fixed-budget (Top-k) retrieval or rely on proxy scores that are computationally expensive and biased. To address these limitations, we propose RaBitQCache, a novel sparse attention framework that utilizes randomized rotated binary quantization and high-throughput binary-INT4 arithmetic to efficiently estimate attention weights. Our proxy score serves as an unbiased estimator with a proven error bound, enabling adaptive Top-p retrieval that dynamically adjusts the token budget based on actual attention sparsity. We further implement a hardware-aware system with asynchronous pipelining and lazy updates to mask overhead. Evaluations demonstrate that RaBitQCache significantly accelerates inference and reduces memory I/O while preserving generation quality compared to state-of-the-art baselines. Code is available at https://github.com/Sakuraaa0/RaBitQCache.git.

Large language models increasingly operate over long contexts, where the KV cache becomes a dominant memory bottleneck: its size grows linearly with sequence length and must be retained throughout decoding, making full GPU caching prohibitively expensive without compression. Existing KV cache compression methods struggle to balance efficiency with faithful context preservation. Token eviction discards information, while semantic grouping fixes compression decisions at prefill time; neither can recover token-level detail from a compressed span once it becomes relevant during generation. As a solution, we propose SeKV, a resolution-adaptive semantic KV cache that organizes context into entropy-guided semantic spans and stores them across a GPU-CPU memory hierarchy without discarding information. Each span keeps a lightweight summary vector on GPU for coarse routing and a low-rank SVD basis on CPU for on-demand token-level reconstruction. A trained zoom-in mechanism selectively expands query-relevant spans during decoding, enabling precise retrieval without materializing the full KV cache on GPU. SeKV enables adaptive token-level reconstruction while keeping the base LLM fully frozen and adding fewer than 0.05% trainable parameters. Across four benchmarks, SeKV improves over the strongest semantic compression baseline by 5.9% on average while reducing GPU memory by 53.3% versus full KV caching at 128K context. Code is available on https://github.com/AmirAbaskohi/SeKV.

A stateless inference server (vLLM, SGLang, TensorRT-LLM) idles between requests while the accelerator waits; a stateful session reclaims that idle time. Speculative pre-positioning decodes the session forward to its next decision point with the target model's own forward pass and no draft model, moving the cross-request prefill and entry-decode off the critical path: the next request resumes from a pre-paid entry on its delta, or, when a confidence gate fires, is answered from a cached distribution in one near-constant vocabulary scan with no decode, at a cost only of energy and a rare, bounded false accept. The payoff is conditional on capability: a capable model fires the gate at near-full coverage and about 87% precision (a smaller one never clears it), returning the first token in about 1.0 ms versus the 39 ms decode a prefix cache still pays.

Large language models (LLMs) excel at complex tasks like question answering and summarization, thanks to their ability to handle long-context inputs. However, deploying LLMs is costly, not only due to the high computational demands of quadratic complexity of self-attention and auto-regressive generation, but also because of the significant memory overhead required for storing the key-value (KV) cache during inference. To reduce the memory cost, existing KV-cache eviction strategies leverage the sparsity in attention to selectively store a subset of tokens. While reducing the memory footprint, such approaches show a considerable drop in performance, especially in tasks that require long-context reasoning. We identify that the drop in performance is linked to a reduction in the coverage of unique tokens. Additionally, we theoretically show that reduced coverage limits the mutual information between inputs and outputs, thereby impairing predictive accuracy. To this end, we introduce K-VEC, a novel coverage-aware KV-cache eviction strategy that prioritizes token coverage while evicting tokens in the cache. K-VEC introduces a cross-head and a cross-layer coverage module to enhance token retention across attention heads and model layers, mitigating performance degradation caused by low coverage. Evaluated on 16 LongBench subsets, K-VEC exhibit up to 10.35 points improvement over the existing methods under the same eviction rate and memory constraint. Comprehensive evaluations validate the effectiveness of our approach and demonstrate its potential for efficient LLM deployment in resource-constrained settings.

Diffusion language models (DLMs) have recently emerged as a promising alternative to conventional autoregressive language models. By generating multiple tokens in parallel during each denoising step, they offer higher inference throughput while maintaining competitive quality. However, realizing these throughput gains while meeting latency SLOs in a serving system requires addressing challenges introduced by DLMs' unique characteristics. These include navigating the speed-quality tradeoff created by confidence-based denoising, choosing appropriate parallelization levels across model instances under fluctuating load, and coordinating approximate KV caching mechanisms that introduce non-uniform per-step costs. To address these challenges, we present DiLaServe, a cluster-level serving system for DLMs. DiLaServe enables deadline-aware scheduling and adaptive load control through confidence-threshold adjustment, and dynamically reconfigures the cluster by solving a quality-aware optimization problem, while explicitly modeling the step-level heterogeneity introduced by approximate KV caching. Across multiple benchmarks and real-world traces, DiLaServe improves SLO attainment by up to 56.6 percentage points and reduces end-to-end request latency by up to 46% while incurring less than 1% accuracy drop.

LLM agents can share more than text. In some systems, an agent can send a short visible message while also passing its full KV-cache state to another model. This hidden state can help the final model combine evidence from several agents, but it is also hard to inspect. A visible message may look harmless even if the hidden state has been changed. We study this problem in a multi-agent question-answering setup. Specialists each see part of the evidence, send a short commitment, and pass full KV-cache state to a coordinator. In clean runs, this latent collaboration improves over a matched text-only version. On transformed HiddenBench with Qwen3-4B, it reaches EM/F1 of 0.338/0.486, compared with 0.231/0.369 for text collaboration. Qwen3-8B and HotPotQA runs show the same direction of improvement. The problem appears when one specialist is malicious. Some false visible commitments can steer answers. More seriously, changing the hidden KV state can collapse performance even when the visible commitment still looks plausible. A verifier that checks only text misses this failure mode. Simple magnitude checks catch some obvious corruptions, but adaptive attacks can evade them while still damaging the final answer. The most reliable fix we find is not to guess whether hidden state looks normal, but to protect it in transport. We implement an HMAC-SHA256 manifest that binds the specialist, session, model, visible commitment, tensor metadata, and payload digest. It accepts all 774 honest replayed payloads and rejects all 295 recorded tampered payloads. The main lesson is that full-KV latent memory can be useful, but it should be treated as a security-sensitive object, not as ordinary internal model state.

Long-context LLM inference faces a fundamental conflict: head-adaptive compression algorithms (e.g., Top-$p$ nucleus sampling) offer superior accuracy by dynamically fluctuating memory budgets, yet modern inference engines (e.g., vLLM) demand rigid, static memory patterns to leverage CUDA Graphs and PagedAttention. We resolve this ``Static-Dynamic'' mismatch with HARD-KV, a unified framework that that bridges dynamic selection with rigid system constraints. HARD-KV introduces a Cascade Cache hierarchy, managing the token lifecycle across dense, sparse, and condensed tiers. Crucially, we propose a Logits Calibration mechanism that normalizes diverse importance metrics into a unified probability space, enabling consistent Top-$p$ budgeting across heterogeneous heads. To bridge the efficiency gap, we offer a system-level solution, which rewrites fragmented, dynamic indices into contiguous physical layouts compatible with high-performance inference engine. Extensive experiments on math-reasoning benchmarks (AIME, U-Math) verify that HARD-KV achieves up to 2$\times$ throughput improvement over static baselines while maintaining high-fidelity generation in 10k+ token scenarios. Code is available at https://github.com/SuDIS-ZJU/HARDInfer.

Model Predictive Control (MPC) is the standard predictive layer in hierarchical quadruped controllers, but the per-cycle QP solve limits the update rate achievable on embedded processors. Because legged gaits revisit a bounded region of state space, MPC solutions admit caching and reuse. This paper proposes \emph{Certified CacheMPC}: a Locality-Sensitive-Hashed cache of horizon contact-force trajectories, partitioned by contact mode, retrieved at query time and accepted only when an a-posteriori per-query certificate confirms primal feasibility and a Lagrangian dual-gap upper bound on cost suboptimality. A bounded-budget controller schedule combines top-$K$ certified retrieval, a deadline-bounded QP solve, and a shifted last-certified fallback. The framework is evaluated on a Unitree Go2 across $2{,}038$ usable cold-controller MuJoCo trials, including a $600$-trial $n\!=\!50$ campaign at three failure-boundary cells, and a first-deploy session on the on-robot NVIDIA Orin NX. The un-gated cache delivers a $25\times$ median solve-time speedup in simulation and an $18.7\times$ median speedup on hardware. At $n\!=\!50$ no statistically significant difference in closed-loop stable rate is detected between the cache variants and the no-cache baseline at any tested cell. The certificate's contribution to closed-loop safety is not resolvable at the present sample size.

Reasoning capability has advanced rapidly in large language models (LLMs), leading to an increasing size of key-value (KV) cache in both prefilling and decoding stages. Existing KV cache compression methods mainly rely on attention weights to estimate token importance. While attention effectively captures contextual relevance, it overlooks complementary information-theoretic signals related to predictive uncertainty and token informativeness. In this paper, we revisit token importance from a forward-looking perspective and introduce \textit{Forward Influence}, a metric that measures how compressed tokens affect future contexts. Our analysis reveals that tokens selected by attention scores mainly influence nearby contexts, whereas tokens associated with high predictive uncertainty exhibit substantially stronger influence on distant future contexts. Based on the observation, we propose \textbf{InfoKV}, an entropy-aware KV cache compression framework that incorporates information-theoretic signals. It combines token-level predictive uncertainty with layer-wise representation evolution and integrates the resulting entropy scores with attention scores during reasoning. Experiments on long-context reasoning benchmarks with Llama-3.1, Llama-3.2, and DeepSeek-R1 demonstrate that InfoKV consistently outperforms existing attention-based KV compression methods in both long prefilling and decoding scenarios.

Diffusion Transformers (DiTs) have driven substantial progress in image and video generation but suffer from prohibitive computational costs. Feature caching accelerates inference by reusing intermediate representations. Existing methods rely on historical features for implementation simplicity, yet suffer from severe error accumulation at high acceleration ratios. To address this limitation, we investigate the nature of the requisite feature correction. We demonstrate that the optimal calibration update is characterized by a shared low-rank subspace across diverse prompts. Guided by this structural insight, we propose LearniBridge, a learnable calibration mechanism for feature caching that bridges multiple timesteps through lightweight LoRA updates. This mechanism enables effective calibration requiring only 3-5 training samples. Extensive experiments on image and video generation show that LearniBridge achieves up to $5.87\times$, $5.75\times$, and $4.10\times$ acceleration on FLUX, HunyuanVideo, and WAN2.1, respectively. On WAN2.1, it improves VBench by 1.28% over the previous SOTA at $4.10\times$ acceleration. Our code is available at https://github.com/Iiiiiiirene/LearniBridge.

Streaming video understanding (SVU) must answer queries that arrive asynchronously while visual tokens stream continuously under strict GPU-memory and query-time latency budgets. A key challenge is delayed query: decisive cues may appear briefly, yet many subsequent updates occur before the query arrives, increasing the risk that those cues are evicted or diluted under bounded memory. We propose ProtoKV, a constant-footprint SVU memory that represents far history as a fixed-capacity summary state rather than retaining token instances. ProtoKV keeps an exact near-window KV cache and aggregates older content into a semantic-spatial prototype bank with residual statistics. At query time, each prototype is exposed through a bounded pseudo-token interface that is drop-in compatible with standard attention. Under matched budgets and comparable query-time cost, ProtoKV improves accuracy by up to 12.5 points over token-retention baselines on SVU benchmarks in the long-delay regime, with gains that grow as query delay increases.

Interleaved multimodal reasoning improves visual grounding by revisiting visual evidence during multi-step generation, yet existing methods typically rely on token replay, repeatedly forwarding selected visual tokens. A natural shortcut is to reuse the historical visual key-value (KV) cache directly. However, we identify a critical failure mode of this strategy: cached visual keys are already bound to their original positional context. Such stale positional binding distorts attention under later decoding contexts and can trigger severe autoregressive decoding collapse. This failure suggests that effective cache reuse requires reconstructing visual evidence under positions compatible with the current decoding state, rather than directly copying position-bound historical cache entries. To this end, we propose Position Rebinding Cache Reuse (PRCR), a cache-level framework for replay-free visual revisiting. PRCR stores raw visual KV cache together with their original spatial coordinates, then reassigns position-compatible coordinates to select entries and rebinds their keys before injecting the reconstructed cache into the active decoder cache. This design reuses historical visual evidence while preserving textual positional continuity and relative visual structure. Experiments across multiple multimodal reasoning benchmarks show that PRCR achieves replay-level or better performance, improving average accuracy by 5 percent and reducing visual-revisiting computation by up to tens of thousands of times.

As reasoning models emit chains of thought tens of thousands of tokens long, KV cache increasingly becomes a deployment bottleneck. Existing cache eviction methods rank tokens by attention weight, which is a noisy importance proxy in long reasoning traces, and prohibits the use of fused kernels in production inference by forcing the model to materialize the attention matrix. In this work, we instead score tokens with a metric we term the epiphany score: the change in the model's internal representation, read directly from the forward pass with no attention matrix and negligible extra state. Our resulting cache eviction method, EpiKV, requires no training, classifier, or custom kernel, and can be used directly in FlashAttention inference stacks unchanged -- scaling to a 16x longer feasible context than attention-based scoring. upper-mid layers negatively) and remove a positional trend with a causal rolling z-score. At a 4096-token cache EpiKV reaches 72% on MATH-500, matching the strongest attention-based baseline (ThinKV 71%, H2O 67%); a lag-normalized KV variant reaches 37% on AIME-2024 at 8192 tokens against the best of them (33%), at up to 2.8x the speed.

Emerging LLM services increasingly host many sparse MoE models, yet most models receive sparse requests and remain cold. This creates a GPU memory problem: model weights are stable and model-determined, while KV-cache is transient and demand-determined. Because cold models rarely reach peak KV-cache demand at the same time, reserving worst-case KV capacity per model wastes memory; a shared KV-cache pool can instead provision aggregate active demand. However, KV-cache sharing is not sufficient when weights and KV-cache remain in a monolithic GPU memory pool. Static weights compete with dynamic KV-cache, and KV-head-limited attention under cold, low-concurrency traffic exposes only a fraction of replicated KV capacity, leading to low GPU memory utilization and weak long-context support. We present CrossPool, a serving engine for cold MoE models that separates FFN weights and KV-cache into two GPU memory pools: a weights pool that consolidates FFN weights across cold models, and a KV-cache pool that dynamically serves active requests while keeping attention local to KV-cache. CrossPool combines a KV-cache planner and virtualizer, a layer-wise pipeline scheduler that hides hidden-state transfers, and persistent kernels with control lowering to reduce CPU-GPU control overhead. With efficient GPU memory pooling, CrossPool underpins bursty long-context requests and outperforms the state-of-the-art kvcached-based multi-LLM serving system, reducing P99 TBT by up to 10.4x.

Long-context large language model (LLM) inference is increasingly constrained by the memory footprint and decoding cost of key-value (KV) caches, limiting sustainable deployment on resource-constrained hardware. Existing KV cache eviction methods typically apply heuristic token scoring over all heads in GQA-based LLMs. These methods ignore the different functionalities of attention heads, leading to the eviction of critical tokens and thus degrading the performance of LLMs. To address this issue, we propose CompressKV, a resource-efficient KV-cache compression framework for GQA-based LLMs. Instead of aggregating attention scores from all heads, CompressKV identifies Semantic Retrieval Heads (SRHs) that capture both the initial and final tokens of a prompt and semantically important mid-context evidence, and uses them to select tokens whose KV pairs should be retained. Furthermore, CompressKV allocates cache budgets across layers according to offline estimates of layer-wise eviction error. Experiments on LongBench and Needle-in-a-Haystack show that CompressKV consistently outperforms existing KV-cache eviction methods across memory budgets. Notably, it preserves over 97% of full-cache performance using only 3% of the KV cache on LongBench question-answering tasks and achieves 90% accuracy with just 0.7% KV storage on Needle-in-a-Haystack. These results demonstrate an improved resource--performance trade-off for long-context LLM inference. Our code is publicly available at: https://github.com/TUDa-HWAI/CompressKV

Existing low-bit KV-cache quantizers often treat each cached key as a flat vector. Under RoPE, however, a key's contribution to a future attention logit decomposes into a position-dependent sum over two-dimensional frequency blocks. This makes key-cache quantization a block-wise bit-allocation problem: high-energy RoPE blocks are more sensitive to quantization error and should receive more bits. We introduce Block-GTQ, a RoPE-aware bit allocator for key-cache quantization built on TurboQuant-MSE(TQ-MSE). For each layer and KV head, Block-GTQ computes a label-free energy score for each RoPE block and greedily allocates integer bit widths by marginal gain. Under matched K/V bit budgets, Block-GTQ better preserves RoPE query-key logits on a ten-model diagnostic panel, cutting per-layer MAE by 32-80% at 2 and 3 b/dim K-only quantization and winning all 367/367 layer comparisons against uniform TQ-MSE. These fidelity gains translate to stronger downstream long-context retrieval, understanding, and reasoning. At K2V2 on Llama-3.1-8B-Instruct, Block-GTQ raises the six-task NIAH average from 70.6 to 97.4, and the LongBench-EN average from 36.87 to 53.31. On AIME 2024/2025 with DeepSeek-R1-Distill-Qwen-7B, without an fp16 recent-key buffer, Block-GTQ at K3V2 scores 51.7/37.5, close to fp16's 54.2/37.9, whereas uniform TQ-MSE collapses to 0.0/0.0. We further implement a packed-cache serving path. On a single H800 GPU with Qwen2.5-3B-Instruct, packed K3V3 achieves 3.24x KV-cache compression with fp16-comparable quality, runs 1.34x faster than fp16 FlashAttention2 at 128K context, reduces peak memory from 56.31 GB to 19.85 GB, and remains feasible at 256K and 512K where fp16 OOMs. Code is available at https://github.com/JIA-Lab-research/blockgtq.