Cache

KV-cache compression reduces long-context memory, but aggregate task scores reveal neither which correct executions fail nor why. We present KVDiagnosis, a diagnostic dataset and benchmark with three contributions. First, a 25-method taxonomy groups methods into five mechanism families and links them to eight verified implementations and their valid diagnostic measurements. Second, for every supported method setting, we evaluate all sources in each fixed split against a per-source FullCache control before selecting FullCache-correct/compressed-wrong (C-to-W) rows separately for each method-setting, so no compressor defines another's test set. Third, a common record format links paired outputs and run metadata to cache, likelihood, attention, and decoding measurements with explicit applicability states. On Qwen3-8B, four evidence-aware workloads yield 59 800 supported compressed runs over 2600 sources and 12 520 C-to-W rows. Under fixed diagnostic rules, 63.2% have low or partial measured/projected coverage. Only 19 rows (0.2%) combine high measured/projected coverage with strong likelihood drift; another 2,126 (17.0%) preserve structural position addressability, for which representation fidelity remains unknown, while showing the same drift. Against C-to-C success controls, all ten diagnostics separate failed from successful compression (stratified AUROC 0.684-0.871). Among 96 reproducible low-EAR failures, a controlled 4x evidence-attention boost repairs 29.2%, versus 6.3% under a count-matched sham intervention and 3.3% degradation on matched C-to-C controls. Code and data are available at https://github.com/ChosenQC/KVDiagnosis.

Diffusion caching is a lightweight strategy that accelerates Diffusion Transformers (DiTs) by reusing intermediate features across denoising steps, but existing paradigms face a fundamental trade-off: online heuristics lack global budget awareness, whereas static schedules lack instance adaptivity and fail to flexibly adapt to varying runtime budget constraints. To bridge this gap, we present BAG (Budget-Aware Gating), a novel caching policy that unifies global budget pacing with dynamic, instance-adaptive feature reuse. Rather than relying on hand-crafted rules, BAG employs a lightweight gating network that dynamically decides whether to execute a full computation or reuse cached features at each step by jointly conditioning on the budget state and local trajectory feedback. We train this policy via offline-to-online schedule distillation, transferring the decision-making of offline-searched schedules into a compact online gate. Extensive experiments on FLUX.1-dev and Wan2.1 demonstrate that BAG consistently outperforms state-of-the-art caching methods across various speedup tiers while remaining robust across different resolutions, seeds, and guidance scales. Code will be released.

The key-value (KV) cache is the primary throughput optimization in modern large language model (LLM) inference, enabling prefix reuse across requests. In multi-tenant deployments this cache is shared across tenants, creating a timing side channel: an adversarial tenant can reconstruct another tenant's private prompt by probing cache-hit latency. Three published attacks exploit it -- PROMPTPEEK, EarlyBird and InputSnatch -- reaching up to 100% attack success rate against unprotected vLLM and SGLang, with rates varying by cache architecture and prompt structure. We present KVGov, a governance layer addressing all three attack families' prefix-cache paths under one mechanism. A per-principal salt sigma_p = HMAC_K(secret, principal_id) seeds the block-hash chain, making cache keys cryptographically disjoint across principals. An ablation (N=1000 trials, seed 2026, deterministic judges) isolates this salt as the necessary and sufficient component. KVGov adds ORIGAMI, a Stackelberg water-filling audit scheduler that reduces adversary expected utility by 12.6% at realistic tenant heterogeneity (Gini 0.63), and an evolutionary stability analysis giving a 31.6% adversary-prevalence tipping point below which global caching remains stable. On real hardware (Qwen2.5-7B-Instruct, vLLM 0.26.0, NVIDIA A100) we measure a gate-verified cold/cached TTFT ratio of 0.22, confirming the channel is exploitable at production scale; the defense itself is evaluated in simulation calibrated to those measurements. We replicate the channel on an independent stack (llama.cpp on Apple Metal, ratio 0.093). Finally, isolation and cache efficiency need not conflict: identifying information resides only where prompts diverge, so injecting the salt at that boundary rather than the chain root retains an estimated 93% of the prefix-cache benefit with no cross-principal signal.

Transformer-based large language models (LLMs) achieve strong performance across many tasks, but their Key-Value (KV) cache grows linearly with sequence length, creating a severe memory bottleneck for long-context inference. Existing heuristic eviction methods (e.g., H$_2$O and SnapKV) rely on static attention or positional signals that often fail to capture a token's future predictive influence. We propose DistillCache, a reinforcement learning framework that formulates KV-cache eviction as a sequential decision problem. DistillCache learns a lightweight policy network using rich internal model signals (attention statistics, value norms, entropy, and position) and trains it with REINFORCE via a per-step KL-divergence reward to preserve the full-cache output distribution. On a 7B-parameter instruction-tuned Transformer (Mistral-7B-Instruct-v0.3), DistillCache retains 94.2% of full-cache accuracy on LongBench at a 25% cache budget, outperforming both strong heuristic baselines (H$_2$O, SnapKV) by up to 2.7 absolute points and, under our re-implementations, concurrent RL-based methods (ForesightKV, RLKV) by up to 1.4 points on long-context tasks. On reasoning benchmarks, DistillCache is competitive with the best concurrent method and surpasses it under aggressive compression. It also delivers up to 2.1x full-cache throughput while maintaining competitive practical efficiency. These results highlight the effectiveness of learned, distribution-aware policies for memory-efficient long-context LLM inference.

Long-context LLM inference is bottlenecked by KV cache memory, yet distributing a limited cache budget across layers remains challenging. Existing methods rely on proxies such as layer depth, attention statistics, or representation change. These proxies do not measure how perturbations at each layer propagate to the output and may therefore cause sensitive layers to be underallocated while tolerant layers are overallocated. To address this issue, we propose RippleKV, which allocates cache across layers by estimating how perturbations to each layer's value cache affect the final predictive distribution. RippleKV independently injects norm-adaptive perturbations into each layer's value cache and measures the induced KL divergence at the model output over a small calibration set. Averaging these responses yields a sensitivity profile specific to the model that need not vary monotonically with depth. RippleKV then converts the sensitivity profile into layer budget multipliers by normalizing the sensitivity scores and applying an exponential mapping. A ratio parameter controls the allocation disparity between sensitive and tolerant layers, while a final normalization preserves the KV cache budget. Experiments on LongBench demonstrate that RippleKV achieves the highest average performance among the evaluated KV cache compression methods under matched cache budgets.

Recent advancements in Speech Large Language Models have demonstrated remarkable capabilities in understanding complex audio tasks. Despite this progress, their long-context inference remains severely bottlenecked by prohibitive KV cache memory demands. Existing text-centric compression methods struggle here, often disrupting speech continuity or discarding crucial semantic cues. To address this, we propose VoxZip, a train-free, two-stage semantic-anchored KV cache compression framework. The first stage uses automatic speech recognition (ASR) transcriptions as explicit semantic anchors to temporally align, compress, and fuse audio tokens, significantly reducing the initial KV cache while elevating token information density. To further improve the compression ratio, the second stage employs a dynamic filtering strategy based on temporally decayed accumulated attention to evict non-essential tokens while mitigating early-token bias. Comprehensive evaluations on Qwen3-Omni across six diverse audio benchmarks demonstrate the superiority of our approach. VoxZip excels in long-audio reasoning and consistently maintains high-fidelity perception on short-form tasks. Notably, it sustains over 90% of the uncompressed baseline performance even under an aggressive 20x KV cache compression in long-context scenarios. Furthermore, at a 4x compression ratio, VoxZip yields a 1.9x increase in inference throughput alongside a 3.3x reduction in peak memory overhead. Code and models will be available at https://github.com/MM-Speech/VoxZip.

Large mixture-of-experts (MoE) language models with 26--120 billion parameters exceed the memory capacity of consumer devices through three simultaneous pressures: resident weight matrices, key-value (KV) cache state that grows linearly with context, and dozens of expert sublayers that must be paged on demand. We present RotaryQuant, a three-axis compression system that addresses all three. Mixed-precision weight quantization assigns bit-widths by architectural role: 4-bit for dense layers, 2-bit for routed experts, and 8-bit for the shared expert whose high activation kurtosis resists aggressive compression. LRU expert offloading pages non-resident experts to disk under genuine memory pressure. The novel axis is IsoQuant, a KV cache compression method that applies a Walsh--Hadamard transform followed by block-diagonal SO(4) rotations to isotropize activation distributions before 3-bit scalar quantization, requiring $O(d \log d)$ operations and 256 stored parameters per head versus $O(d^2)$ and 16{,}384 for dense rotation methods. A fused four-kernel Metal GPU pipeline performs attention directly on packed 3-bit tensors without materializing full-precision KV state---a different execution model, not just a quantization scheme. The combined system fits Gemma 4-26B-A4B and Qwen3-30B-A3B within a 16,GB budget and Nemotron-H 120B within 32,GB, running interactively at 9--19 tok/s with near-zero perplexity degradation ($Δ$PPL $\leq +0.0012$) and 100% retrieval accuracy at 32K context.

Large language models can answer knowledge-intensive questions more reliably when they are grounded with knowledge graphs, but systems such as Think-on-Graph and Reasoning-on-Graph repeatedly query the same graph neighborhoods across different questions. In this work, we study this repeated retrieval in Knowledge Graph Question Answering~(KGQA) workloads and propose KGCache, an in-memory cache for one-hop knowledge graph neighborhoods. KGCache is designed to be compatible with both iterative traversal (ToG) and one shot planning (RoG) KGQA paradigms. KGCache is placed between the KGQA engine and the backend serving the KG, so repeated entity requests can be served from cache instead of issuing new KG queries. We evaluate KGCache on WebQSP and CWQ using LRU, LFU, and a trace-aware Oracle policy. Our analysis shows that both datasets contain substantial entity reuse among starting entities and entities reached during traversal. We also explore semantic caching for similar queries, which shows additional hit-rate gains on WebQSP and needs further accuracy testing on CWQ. Entity caching accelerates KG retrieval by up to $1.91\times$, while semantic-context caching achieves up to $1.06\times$ full-system speedup in the evaluated WebQSP configurations, with each hit being up to $3.73\times$ faster.

Large language models (LLMs) increasingly read long inputs in the agentic era, from whole documents and codebases to conversations across many turns. Their inference memory is then dominated by the key-value (KV) cache, the stored attention keys and values of every token the model has read and generated. Because the cache grows with context length and is re-read in full at every generated token, a longer context means more GPU memory. To reduce this cost, most existing methods compress the KV cache by lowering every stored value to the same low precision, a technique known as quantization. They can push this to nearly two bits per value, but rarely further, because quality drops sharply at this 2-bit cliff: four levels are too few for the cache's outlier-heavy values, where a few large entries consume the levels and collapse the rest into noise. A natural remedy is to spend more bits on the channels (feature dimensions) that matter and fewer on the rest, but the raw cache offers no handle: its channels are strongly correlated, so none stands out as more important. Our analysis shows that this handle appears once the cache is rotated into a coordinate system computed from its own statistics, removing these correlations. There, a small fraction of channels carries almost all the information, and spending the budget on those few is far more accurate than spreading it evenly. Guided by this analysis, we develop SPECTRA, a training-free, drop-in codec that re-encodes the cache into this coordinate system and concentrates the bit budget on the channels that carry the signal. On Llama-3.1-8B and Qwen2.5-7B over long-context benchmarks, SPECTRA is near-lossless at 4x compression, competitive at 8x where uniform quantization has collapsed, and reaches up to 12x, pushing usable compression past the 2-bit cliff so the same GPU holds longer contexts and larger batches.

Mixture-of-Experts (MoE) models have outgrown accelerator memory, and offloading expert weights to host memory is now standard. This makes expert cache management an attractive lever: a policy that raised the hit rate would cut expert traffic per token. Evaluating that is a measurement problem, and we find the measurement fragile. With a trace-driven, event-atomic simulator over three MoE models (40, 64, 128 experts), we isolate three evaluation axes that change conclusions, not just numbers. Replay semantics: under a fused-event traffic contract, an inconsistent per-access replay inflates recency-based policies by 27-29% while leaving frequency-based and static ones within 4%, inverting the policy ranking. Workload contamination: probe sets using one instruction template per category produce verbatim-identical generation prefixes; a matched-pair rendering intervention moves the measured early-window effect by 19.4-31.9 points and reverses which workloads look most cache-friendly. Operating regimes: normalized miss fractions do not transfer across models, so the per-step expert union relative to per-layer capacity must be reported -- yet permuting only the temporal order of an identical event stream moves the offline-optimal gap from 44.9% to 30.8%, so it is not sufficient. Corrected, a stable gap to the offline optimum remains (44.2-45.9% over 13 frozen workload compositions). A forced-admission oracle attributes 84.3-96.6% of it to knowing which resident expert is used furthest in the future. A causal next-use predictor, used as an eviction rule, recovers -11.4% of the gap; it picks an optimal victim 3.4% of the time, against 2.4% for a random resident block and 20.6-22.1% for LRU and LFRU. Our position is narrow: in our evaluated settings a large offline-optimal gap substantially overstates the gains recovered by representative lightweight causal mechanisms.

Multi-turn Reasoning-and-Acting (ReAct) agents accumulate growing trajectories of reasoning, tool calls, and observations. Their key-value (KV) caches grow accordingly, increasing memory use and attention cost during model inference. Existing KV cache compression methods reduce these costs by evicting states with low attention scores. However, low attention in the current turn does not imply future irrelevance, as temporarily inactive information may become important later. Snapshot-based eviction methods therefore do not explicitly distinguish temporarily dormant information from information that appears to have completed its role. In this paper, we present CommitKV, which identifies KV lifecycles through commit transitions. Specifically, CommitKV first divides completed agent events into token pages and compares each eligible page's deletion effect before a tool-call commit and after the commit's returned observation has been incorporated. Based on these paired measurements, CommitKV distinguishes dormant pages from high-to-low completion candidates. It then applies a greedy joint test, accepting candidates for retirement only when their combined post-commit effect remains bounded. Finally, at a later compression checkpoint, accepted pages are excluded, a bounded set of pages awaiting post-commit measurement is protected, and the remaining KV states are retained within the cache budget using the same token indices for keys, values, and absolute positions. These mechanisms ensure that CommitKV can distinguish dormant information from information that has completed its observed role and can be safely removed. Experiments on various benchmarks show that CommitKV reduces agent memory use, accelerates end-to-end inference, and achieves higher accuracy than existing KV cache compression methods.

Recent optimization studies on Retrieval-Augmented Generation (RAG) have exploited chunk-level KV cache reuse to avoid processing long retrieved contexts for higher efficiency, while significant information redundancy and noise still remain in the coarse-grained chunks. This paper optimizes the Pareto frontier under low prefill latency constraints while maximizing accuracy by proposing CoinRAG (Contextualized Information Nugget KV Cache Reuse for Long-Context RAG). The name metaphorically reflects our core mechanism: much like assembling small tokens (or "coins") to accumulate a larger value, CoinRAG compositionally reuses offline-computed, fine-grained nugget caches to form a learned contextual representation efficiently in a more semantically relevant but compact manner. Specifically, instead of full-chunk encoding, CoinRAG identifies query-relevant semantic units within retrieved chunks through two-stage retrieval and seamlessly assembles their sliced KV representations with a chunk-level context. Extensive evaluations on LongBench multi-hop question answering tasks demonstrate that CoinRAG significantly reduces operational costs and outperforms the other baselines with a new Pareto frontier and an average 5.3% relative improvement in answer quality (F1) under a standard fast prefill latency budget.

Agentic coding faces growing problems of affordability and wasted tokens. We introduce Blast Radius, a predictive memory management layer that estimates an incoming prompt's reach through coupled context and code channels. NECROPHORESIS enables reversible eviction by archiving dead context verbatim, while Recurring Dead Matter (RDM) identifies and buries repeatedly occurring transcripts. We formulate reversible context eviction over a Polish context space, providing a measurable foundation for retention, recurrence, and eviction while connecting context entropy to resurrection probability. Across seven OpenAI models, Blast Radius reduced token consumption by 17-26%, achieved the lowest overflow rate among tested policies, and remained byte exact reversible. Of 450 buried bodies, 378 were recurring dead matter and zero were recalled. Blast Radius operates beneath HCRC, determining which records to bury and how far an incoming prompt may reach into the codebase. This work contributes to the broader goal of Algosophy: making large language models and agentic coding more reusable and sustainable.

As large language models (LLMs) process increasingly long contexts, KV cache storage and repeated access have become a major bottleneck. Existing KV cache compression methods rely on predefined, fixed compression rules and are typically developed around either token eviction or merging. As a result, cache resources can neither flow freely across layers, heads, and context slots, nor be jointly allocated to balance local resolution and information coverage. Therefore, we propose GraceKV, a global approach for the allocation of resolution and coverage in KV cache compression, and formulate the compression process as a global resource allocation problem under a fixed cache budget. GraceKV treats each layer-KV head-slot combination as an atomic unit and builds a prototype tree. Leaf nodes correspond to token-level KV entries, while each internal node uses a single prototype to compress the KV space covered by its children. A set of non-overlapping nodes in the tree forms the representation of an atomic unit. Adding the root of a new tree expands information coverage, whereas splitting a selected node improves local resolution. All candidate actions compete globally for a shared cache budget. Finally, the nodes retained across all trees form the compressed KV cache. This process adaptively determines the allocation of cache resources among atomic units globally and the balance between resolution and coverage. GraceKV requires no additional training, and the entire compression and inference process is performed on the GPU. Systematic experiments across diverse long-context tasks and compression ratios show that GraceKV ranks first in 24 of 32 settings and remains robust up to 128-fold compression. These results validate the effectiveness of global budget allocation in coordinating information coverage and local resolution.

Facial expression recognition (FER) in videos is challenging because models must identify subtle, temporally evolving affective states that vary across individuals. Although vision-language models provide transferable visual-semantic representations, models trained on subject-independent data often degrade under subject-specific distribution shifts at inference time. Existing test-time adaptation (TTA) methods commonly update model parameters during inference, increasing computational cost and latency. Cache-based methods avoid parameter updates, but they usually require enough target samples to form reliable class prototypes, which is difficult early in adaptation and for rarely observed classes. We introduce Energy-Based Cache Personalization (EB-CaP), a subject-based online TTA method for video FER that generates class-specific prototypes personalized to each target video. EB-CaP uses a lightweight energy-based model to sample prototypes from the current unlabeled video and populate a personalized cache online, without accumulating large amounts of target data or storing diverse source prototypes. Its energy function relies only on pretrained CLIP: similarities between the target video embedding and class text embeddings guide prototype sampling. In parallel, positive and negative caches store reliable and uncertain target embeddings. An adaptive entropy gate controls cache updates according to the evolving confidence distribution, while a diversity gate limits redundant samples. Final predictions combine cache-derived scores with the current CLIP scores. Experiments on BioVid, StressID, and BAH show that EB-CaP outperforms state-of-the-art TTA methods while maintaining low computational and memory overhead. Code is available at https://github.com/MasoumehSharafi/EB-CaP.

Modern models no longer keep a plain KV cache: latent caches, learned sparse selectors and recurrent states each carry the model's memory in a different form, and each fails differently under compression. We give a runtime observability contract that covers all four memory classes with three operators, instantiate it on six model configurations across five architecture families, and compose the per-stage bounds into an executable request-level risk ledger. Contracts carry their error metric as a type -- composition is only defined when metrics match, and this check rejected our own first composed chain; the repaired chain crosses metrics through two proved bridges, and whatever no formal system can certify is measured instead, dropping the composed tier to empirical automatically: every claim is certified, partially certified, or empirical, composition inherits the weakest tier, and the tier is decided by the machine. Replayed over $12.4$M entry reads and run under eight-way concurrency with per-request budgets and fail-closed identity attribution, the ledger quantifies the honest trade-off on today's witness and holds its risk budget with zero violations. A fused always-on probe observes a declared one-layer subset under CUDA graphs inside the serving noise floor. Applied to a served DeepSeek-V4 stack with a packed compressed-KV prototype, the same machinery localizes a silent corruption to a precise structural boundary -- exact in the eviction-free, identity-isolated regime, with every observed failure in an eviction or slot-reuse regime -- through a machine-adjudicated discrimination campaign whose calculus rejected two of our own confounded inferences along the way. All artifacts, guards, and the Lean development are released at https://github.com/metask-ai/witprobe-attention-memory; every number in this paper regenerates from the shipped artifacts by one command.

Federated learning repeatedly incurs local optimization and model-update transmission. We study DG-FedReuse, a simulator-level mechanism that allows selected clients to contribute age-decayed cached updates when a stochastic head-gradient discrepancy proxy remains below a round-dependent threshold. A hard cache-age limit and minimum fresh-client quota constrain reuse, while fresh updates use an adaptive per-tensor Top-K numerical-field representation. Experiments cover six image-classification datasets, 50 virtual clients, Dirichlet label heterogeneity (α=0.5), and three seeds. At a common 90-round budget, DG-FedReuse yields 83.36-85.42% modeled update-data-field uplink saving, compared with 76.88% for matched Top-K FedAvg; the seed-aligned accuracy differences range from -5.29 to -0.14 percentage points. Best-observed test accuracies obtained under test-controlled checkpointing are retained only as exploratory archival evidence and range from -2.38 to +0.45 percentage points relative to matched FedAvg. A symmetric dense-model-downlink sensitivity reduces the headline saving to 41.68-F42.71% and the incremental gain over Top-K FedAvg to 3.24-4.27 percentage points, demonstrating the dependence of communication conclusions on the accounting boundary. The study characterizes the proposed reuse rule in the implemented simulator; it does not establish unbiased generalization, end-to-end bandwidth reduction, runtime or energy savings, faster convergence, or superiority over existing stale-update and lazy-aggregation methods.

Autoregressive large language model inference is increasingly constrained by the memory footprint of the Key-Value (KV) cache. A dominant line of work reduces this footprint by evicting tokens that appear unimportant under attention-derived scores. However, such policies make an implicit irreversible decision: once a token is evicted, it cannot become useful again. We show that this assumption is brittle during decoding. Token and window importance drift as generated queries evolve, causing standard eviction policies to permanently discard states that later receive substantial attention under the full-cache model. To characterize this behaviour, we introduce Future Missed Mass and Global LIR, two diagnostics that measure future attention assigned to discarded states and the reactivation of historically inactive regions. We propose QEvict, a three-tier KV-cache management scheme that replaces binary retain-or-delete eviction with recoverable eviction. QEvict maintains high-confidence windows in full precision, stores intermediate windows in a quantized recoverable tier, and deletes only the lowest-confidence windows. During decoding, cumulative attention scores update window importance and when a quantized window becomes important again, it is dequantized and promoted to the full-precision. Under a fixed memory budget, this design preserves broader historical context while retaining exact full precision for the most important regions. Across long-context understanding, retrieval, and reasoning benchmarks, QEvict consistently improves over representative eviction and quantization baselines, reducing missed attention and improving information retention

Multi-agent LLM systems relay key--value caches instead of text and credit their gains to exchanged ``latent thoughts''. That credit is a claim about \emph{which} example's cache is relayed, not merely that one is. We audit it causally in released systems. The cache is replaced with deranged (mismatched-example), zeroed, and moment-matched random counterparts, under two regimes defined by whether the receiver needs the sender's private information. Where it does, the battery reads ceiling: 100% against 23--25% for answer-irrelevant relays on the primary backbone, a contrast replicated across three families, five checkpoints, and a prose document-QA surface. Where it does not, a pre-registered five-seed protocol establishes equivalence within 2.8 points, a margin anchored to the audited system's reported gain, under Holm-corrected TOST on GSM8K and ARC-Challenge across three Qwen3 scales and on MedQA at 8B (one cell shows a small detected advantage inside the margin); a second family shows no detected advantage. A large cache effect need not be a pairing effect. In one natural cell, zeroing the relay costs 14.7 points; a mismatched cache, 0.4. Nor is need sufficient: under the same test, delivered channels span ceiling (LatentMAS's native relay), partial (KVComm's layer subset), and no detected example-specific transfer (C2C's released projector). Benchmark deltas do not by themselves establish latent-thought transmission; establishing it takes a mismatched-cache audit, which we release.

Large Reasoning Models (LRMs) excel on complex tasks through long chain-of-thought (CoT) reasoning, but their lengthy intermediate steps cause severe overthinking that inflates inference cost. KV-cache compression is a common solution, yet existing reasoning-oriented methods apply a uniform policy across the trajectory and judge compression only by what it removes from the cache. Two observations point the other way. First, a reasoning state's tolerance to context loss varies along the trajectory, and process reward tracks it: deleting tokens at high-reward steps preserves accuracy far better than deleting the same budget at random. Second, compression is not free on the generation side, since a smaller cache leads the model to generate more tokens, partly canceling the saving. Together these motivate coordinating both sides under a single process reward. We propose ReCo (Reward-Coordinated Compression), a step-wise framework in which a lightweight process-reward estimator scores each completed step and drives three components: (1) reward-adaptive KV-cache compression that shrinks the retained cache harder at high-reward steps and less at low-reward ones, (2) a reward-banded penalty on reflection tokens that curbs redundant generation, and (3) confidence-based early stopping that triggers when the reasoning is reliable. Across three reasoning models and six benchmarks, ReCo reduces generated tokens by 37%-65% and end-to-end latency by 2.08x-2.35x over Full CoT, all while largely preserving accuracy.

Production deployments often swap between different-sized models in a family for cost-quality cascading, mid-conversation switching, and routing, and each swap forces the receiver to repay the prefill from scratch. We propose cross-model KV cache transfer, where the receiver reuses the source's KV cache, skipping prefill. We find that cross-model KV has substantial linear structure across matched-KV pairs, where source and target share KV head count and per-head dimension. On Qwen3 14B->32B, one source layer explains 56% of variance in the target's keys and 32% in values, rising to 79% and 65% with multiple source layers. Building on this, we design a closed-form ridge mapper that operates per head and proceeds in three steps. First, for each target layer we select the top-k most predictive source layers and concatenate their KV as input. Second, we strip RoPE from the keys before mapping, so the fit is position-free and reusable across context lengths. Third, we fit ridge regression on a small calibration set of 500 FineWeb-Edu sequences of 1,024 tokens each. Surprisingly, across six pairs in three families, this linear mapper retains 73-98% of the receiver's standalone-prefill accuracy on four pairs, while two degrade sharply. A nonlinear MLP recovers up to +37 pp HellaSwag retention on the failures. The mapper runs 2.7-25x faster than re-prefill and remains stable across multi-turn handoff, making cross-model KV cache transfer practical.

Long-context LLM decoding reads the key-value (KV) cache at every step. Loading it takes longer than computing attention over it, so throughput is bandwidth-bound. Hence, reducing the cache size can raise both decoding speed and serving capacity. The challenge is to reduce cache size while preserving the attention products, keeping reconstruction cheap, and using a fixed per-token bit count. At two bits per element, the most competitive methods rely on orthogonal transforms. However, existing techniques are either data-oblivious or use the query statistics without deriving the transform from a distortion criterion. Moreover, they rely on transforms built on top of random or Hadamard rotations, which equalize variances across entries rather than compacting energy, and fixed-width scalar quantizers, which are suboptimal at low rates. In this paper, we formulate KV cache quantization as a transform coding problem in which distortion is the error in the attention products. We derive closed-form optimal transforms for keys and values from calibration statistics, under a high-resolution model. We show that the optimal key transform is not orthogonal and satisfies a generalized Parseval relation: the attention-aware distortion becomes mean-squared error (MSE) in the transform domain. Thus, we can use MSE-optimal vector quantizers applied directly to the transformed key coefficients. To meet the fixed-width layout requirement, we show that grouping coefficients into equal-volume partitions makes equal-size codebooks attain the variable-rate optimum under the same high-resolution model. At two bits per element, our method, termed NOVA-KV, recovers most of the long-context retrieval accuracy lost by scalar quantization methods at comparable throughput.

Long-context inference with large language models is constrained by the linear growth of the key-value cache to sequence length. While pruning offers mitigation, prevailing methods determine query-specific token importance that cannot be reused across unseen queries. In contrast, we introduce TaskPress, a framework for task-guided, query-agnostic KV cache eviction. Instead of optimizing the cache for a single query, TaskPress constructs a reusable memory representation conditioned on a high-level task guide. The guide functions as a meta-query during prefill to filter irrelevant tokens before downstream queries are issued. In addition, TaskPress leverages quantization scale factors as a zero-cost signal for detecting influential representation outliers, providing an efficient proxy for token importance. Experiments on conducted on various tasks with long context input demonstrate that TaskPress efficiently creates a compact, reusable cache across diverse queries.

Existing low rank KV cache methods preserve either model weights or key variance, neither of which directly reflects the attention scores used during inference. We derive the expected attention score distortion caused by rank r key compression and show that it yields a covariance weighted low rank objective. Under a margin condition, controlling this distortion also improves top k recall. The optimal rank r solution has a closed form asymmetric factorization obtained from the SVD of the covariance weighted query key operator. This motivates SAKI, a training free KV cache index that directly preserves attention scores rather than key reconstruction quality. Across LLaMA 3.1 8B, Qwen 2.5 7B, Mistral 7B v0.1, and Llama 3.2 3B, SAKI outperforms key PCA at every tested rank. At rank 32, it removes 13 to 30 percent of PCA's remaining top 64 recall error, including improvements from 0.748 to 0.799 on LLaMA 3.1 8B and from 0.786 to 0.850 on Qwen 2.5 7B. It improves 68 to 89 percent of attention heads per model, with the largest gains in deeper layers. Predicted score MSE reductions closely match empirical measurements, with a Pearson correlation of 0.997, while ablation studies confirm that the gains arise from optimizing the attention score objective rather than covariance weighting alone. Analysis of the scoring operator further explains why weight only, invariant subspace, and key reconstruction methods can be suboptimal. SAKI uses random-matrix theory to separate genuine covariance signal from autocorrelated sampling noise, matching PCA with only 512 calibration tokens and adding value exactly where PCA sees no reliable signal.

The key-value (KV) cache is the primary memory bottleneck in long-context LLM inference. Existing approaches attack it from opposite ends: eviction methods permanently discard tokens, degrading performance whenever a discarded token later proves essential, while quantization methods retain all tokens at low precision but offer limited compression. We propose AnchorKV, a compression scheme that shrinks the cache by $20\times$ without discarding a single token. AnchorKV represents the cache using a small set of anchors stored exactly, expresses every other token through its most similar anchor, and refines only those whose approximation most affects the model's output. AnchorKV consistently preserves accuracy across models and datasets, retaining 99% of the full-cache score at the 70B scale, while keeping the entire context at a fraction of its cost.

Audio-driven video generation (A2V) has achieved promising progress in synthesizing temporally coherent and audio-visually aligned videos, yet its inference remains expensive due to the iterative denoising process of diffusion models. Existing caching methods mainly exploit temporal redundancy in visual features while overlooking the cross-modal alignment of A2V, where audio drives visual generation with highly non-uniform temporal importance. In this paper, we identify two levels of misalignment in existing A2V caching methods: temporal-semantic and computation-storage misalignment. To address them, we propose EchoCache, an energy-guided cross-modal caching framework for efficient A2V generation. EchoCache leverages audio time-frequency energy as a saliency anchor to guide latent-level cache updates and further introduces a dynamic timestep-latent caching mechanism with quantized cache management for joint efficiency and memory optimization. Extensive experiments on mainstream A2V models show that EchoCache consistently improves the latency-quality trade-off while preserving generation quality and audio-visual consistency. In particular, on Wan2.2-S2V over the EMTD benchmark, EchoCache achieves a 2.46x speedup with the best overall performance. Code is available at https://github.com/IF-LAB-PKU/EchoCache.

Audio-driven portrait animation has advanced rapidly with diffusion-based generative models, yet real-time one-shot generation with expressive emotion control remains challenging. Existing methods often suffer from insufficient emotion-aware motion priors and expensive appearance computation during multi-step denoising. To address these issues, we propose Proxy Avatar Meets Low-Rank Caching, a cascaded framework for real-time one-shot emotion-controllable portrait animation. Instead of directly generating the target portrait from audio, our method uses a Gaussian-based emotion proxy avatar as a reusable motion generator, which is trained once on a single identity to produce expressive driving videos from audio and emotion labels. Since the proxy avatar only provides motion rather than target appearance or geometry, a large-scale one-shot retargeting model further extracts identity-independent motion from the proxy performance and adapts it to arbitrary target portraits. To improve inference efficiency, we introduce zero-shot appearance reuse with low-rank caching, which caches reference appearance features at the initial denoising step and models subsequent feature variations using lightweight low-rank adapters. Extensive experiments demonstrate that our method achieves stronger emotional expressiveness, better identity-preserving animation, and substantially reduced inference cost, enabling real-time one-shot portrait animation.

The key-value (KV) cache has become a major memory and bandwidth bottleneck in long-context large language model inference, making ultra-low-bit quantization increasingly important. However, existing rotation-based INT2 methods optimize cache statistics or proxy errors before the complete attention readout, even though the model is ultimately affected by the error propagated through attention and the output projection $W_O$. To address this mismatch, we propose \textit{OptR}, an output-aware rotation method that minimizes post-$W_O$ attention-output error. OptR decomposes the post-$W_O$ attention-output error into key- and value-induced terms and learns per-head orthogonal corrections through the full INT2 quantization and attention path. OptR further applies an attention-equivalent key reparameterization to reduce large channel-wise offsets without changing the softmax distribution. Across three models and five reasoning and coding benchmarks, OptR consistently improves both QuaRot and OSCAR and strengthens long-context retrieval, while preserving the paged KV-cache format with negligible inference overhead.

Semantic caching, which reuses responses to semantically similar requests via their embeddings, has seen growing adoption in LLM serving, offering faster responses and reduced costs. Yet existing schemes are fundamentally vulnerable to cache-collision attacks, wherein an adversary pollutes the cache by injecting crafted queries, corrupting responses to subsequent legitimate requests. We present LaCache, a novel semantic caching scheme that addresses this vulnerability through a conceptually simple yet principled redesign. The key insight is that while the adversary has full control over the adversarial query, it has far less control over its response, which must simultaneously satisfy multiple semantic constraints. Rather than checking only the cache hit of a query, LaCache additionally checks the cache hit of its first k (speculatively) decoded tokens. This design yields two concrete benefits. First, it provides formally guaranteed resilience against cache-collision attacks: we prove that it is impossible to craft adversarial queries that simultaneously elicit malicious responses and collide with benign queries. Second, the enriched index supplies additional semantic context for cache retrieval, improving response relevance. Empirical evaluation across diverse LLMs and benchmarks validates both LaCache's security guarantees and efficiency gains, pointing to a promising direction for robust semantic caching.

Sparse attention is widely deployed in long-context serving stacks, yet no framework audits how discarding blocks changes the influence of specific content on model output. We first establish that the phenomenon is real and causal: Block Sparse Flash Attention (BSFA) route replay across four architectures changes output decisions in 13 of 16 cells, with zero identity-replay label flips. We then introduce a dense-calibrated counterfactual audit using matched probe cards---Gold (carrying the correct answer label), Poison (carrying a target wrong label), and Benign (filler only)---under six-layout position symmetry, isolating the sparsification-specific effect. Two patterns compete. Signal concentration: the selector preserves Gold and Poison blocks far above filler-matched Benign blocks (G$\approx$P$\gg$B across all model--task pairs). Integration loss: discarding blocks severs cross-block attention---confirmed by an ablation where isolating the probe block collapses its influence from 4.48 logits to zero. Compression ratio governs the balance: a full sweep from mild ($c=0.25$) to aggressive ($c=0.75$) compression across four model--task pairs reveals that three of four cells move toward stronger sparse amplification at higher compression, with two exhibiting sign reversals. Three independent arms---BSFA route replay, controlled block-top-$k$, and KV-cache eviction---converge: sparsification changes content influence in ways aggregate accuracy cannot detect. We provide an open measurement framework deployable on any model exposing block identities.

Feed-forward 3D Gaussian Splatting enables efficient novel-view synthesis without per-scene optimization, but most existing methods assume a fixed set of context views and process them jointly. This limits their applicability to online scenarios where calibrated views arrive sequentially and the scene must be updated causally. We present \emph{StreamSplat}, a streaming feed-forward 3DGS framework that incrementally maintains a persistent geometry-grounded scene state and decodes it into renderable 3D Gaussians after each input chunk. StreamSplat centers on a \textbf{Voxel-Aligned Causal Cache (VACC)}, which stores historical 3D tokens in a memory-bounded voxel structure so that memory grows with explored scene geometry rather than stream length. To better reuse history during causal prediction, we introduce \textbf{History-Projected Depth Anchoring (HPDA)} to project cached geometry as depth guidance for current cost-volume estimation, and \textbf{Cache-Guided Feature Injection (CGFI)} to inject cached latent evidence into Gaussian-token regression. Experiments on DL3DV, RealEstate10K, and ScanNet show that StreamSplat remains competitive with state-of-the-art feed-forward 3DGS methods under sparse causal inputs, despite not using future views or full-scene context. More importantly, it scales to long input streams with 256, 512, and 1024 views where fixed-view baselines run out of memory, yielding sustained improvements in novel-view synthesis quality as more observations arrive. The code will be made publicly available upon acceptance.

KV cache compression is commonly evaluated by final-answer accuracy, implicitly assuming that preserving the answer also preserves the reasoning that supports it. We test this assumption for large reasoning models and show that it can fail: under compression, correct answers and the validity of their visible supporting rationales can be preserved at different rates. We study this failure with a controlled fixed-trace replay protocol, which holds reasoning content fixed and isolates whether compression preserves usable information from an already available trace. We evaluate ten token-eviction KV compression methods and one quantization method on three models across mathematical reasoning, scientific QA, clinical calculation, and long-context retrieval. We measure final accuracy, answer-chain consistency, and perturbation faithfulness. Across tasks, token-eviction methods can preserve competitive final-answer accuracy while substantially degrading chain support or perturbation faithfulness. We call this the answer-evidence gap. A coverage-preserving quantization control is substantially less affected, suggesting that the failure is tied less to KV memory reduction itself than to losing access to parts of the reasoning trace. Code is available at https://github.com/famous-blue-raincoat/Safe_KV_Compress.

Query-agnostic KV cache eviction compresses a context once and reuses the resulting cache for arbitrary future queries, but performance can collapse under tight budgets. Existing methods primarily improve which original KV pairs are retained. We introduce RestoreKV, which complements this selection-based formulation with learned restoration under the same total KV budget. Our key insight is that, although the information lost through eviction is context-specific, the mechanism for generating its compact complement can be shared across contexts. After context prefill, a few restore tokens attend to the full KV cache in a single LoRA-adapted pass, generating a compact, context-conditioned restore cache. The base importance scorer and eviction rule remain unchanged, and the adapters are disabled for all subsequent queries and decoding. RestoreKV is trained through parameter-efficient self-distillation from the frozen full-cache model, optimizing only $0.4\%$ of the parameters and requiring no task-specific tuning. Across four backbones and four long-context benchmarks, RestoreKV substantially reduces compression-induced degradation. On Qwen3-4B, it improves 59 of 60 paired, budget-matched settings across five base eviction methods; at a $5\%$ budget, it raises KVzip from $38.2$ to $73.2$ on RULER-4K. Applied to KVzip+, RestoreKV reaches $86.4$ RULER accuracy at $16\times$ compression on the KVPress Benchmark, while adding less than $0.5\%$ one-time cache-construction overhead in a 32K-context evaluation. Our project page is available at https://paper.pnu-cvsp.com/RestoreKV/

LLM agents accumulate long trajectories of reasoning steps, tool calls, and environment feedback, making the KV cache a major inference bottleneck. KV cache compaction can reduce this cost, but most prior methods assume a static context where future queries are known or can be approximated offline. Agents instead require online compaction: new information must be compressed before future relevance is known, using proxy queries cheap enough for the inference path. We study online compaction across token eviction (TE) and attention matching (AM), adapting both to compact agent turns and comparing cheap proxy sources such as boundary, repeat-prefill, and delayed future-generation queries. Experiments on BrowseComp-Plus and WideSearch show that immediate compaction often hurts performance, whereas delaying compaction to use the agent's future queries recovers much of the gap. Moreover, TE is often more robust than AM under imperfect proxies. Across models at different scales, TE preserves most of the accuracy while reducing KV cache by 80%, and can improve throughput over the no compaction baseline. These results position proxy-query selection as a core design choice for practical online KV compaction.

The growth of context window lengths in Large Language Models (LLMs) significantly enhances their long-context capabilities but incurs prohibitive memory costs due to the Key-Value (KV) cache. Although low-rank compression of KV cache is a promising remedy, existing methods face a dilemma: offline approaches depend on external calibration data, whereas online approaches incur substantial compute for full-prompt decomposition and reconstruction. In this paper, we propose S$^4$R, which builds low-rank subspaces from selectively sampled tokens and computes attention over a sparsely reconstructed KV representation. S$^4$R uses prompt-aware initialization to build initial key/value bases from a representative prompt subset, trading off calibration-data dependence against prefilling cost. Because fully reconstructing the cache at every decoding step is prohibitively expensive and hurts throughput, we further adopt sparse reconstruction to retain only informative positions during decoding. Extensive experiments on LongBench and RULER with Llama and Qwen model families show that S$^4$R achieves up to 5$\times$ KV compression with near full-cache accuracy, combining the efficiency of fixed compression with the adaptability of prompt-dependent methods.

Token skipping is a widely used training-free way to accelerate vision--language--action (VLA) models by bypassing computation for most visual tokens at each control step according to a gate. When the next gate is harvested from the previous accelerated forward, however, the tokens skipped at one step are also the ones least visible to the next gate, and the damage can compound across control steps until the task fails. We study the two mechanisms this class is built on, reuse and deletion, crossing each against where its gate signal comes from on identical episodes. At a skip ratio of 0.9 on LIBERO-Object, both collapse when the gate comes from the model's own accelerated forwards, to 0.68 under reuse and to 0.31 under deletion against a dense 1.00, and the collapse is invisible to the action-level detectors we evaluate. What separates collapse from dense-level operation is not the mechanism but whether the gate is clean, computed by a forward that skipped nothing. We therefore propose actuation-slack refresh, one dense pass run while the robot executes its current action chunk, off the critical path, that hands the next step a clean gate and a fresh KV base. Since the measured detectors do not reliably reveal the failure, the refresh is unconditional rather than triggered. Both mechanisms then recover to 0.98, keeping the speed of skipping and the information of a dense pass. We then integrate the refresh into state-of-the-art caching and pruning methods across two VLA policies, 4 LIBERO suites, and 4 SIMPLER tasks, where it repairs every collapse caused by using a self-harvested gate. Serve latency drops 18--22% below dense, measured both in simulation and on a physical robot. Where the gate signal comes from, not how tokens are skipped, decides closed-loop reliability for accelerated VLAs.

KV cache compression is essential for efficient long-context inference. Existing eviction methods permanently discard unselected tokens and consequently remove their aggregate contribution to attention. Merging-based alternatives preserve more information but can perturb retained keys and values that should remain exact. We observe that the information omitted by cache eviction can be formulated as residual statistics in both the numerator and denominator of softmax attention. Based on this observation, we propose ResKV, which divides a fixed KV budget into an exact main cache and a compact residual cache that reconstructs the contribution of omitted tokens. ResKV lets main-cache tokens and residual entries participate in the same softmax normalization, so residual entries restore both attention numerator and denominator mass rather than acting as a post-hoc correction. A construction-time validation proxy determines residual allocation for each layer and KV head, while a decode-time dynamic gate adjusts residual contributions for individual queries. Comprehensive evaluations on LongBench and RULER, covering query-aware and query-agnostic settings, multiple backbones, cache budgets, and representative compression baselines, demonstrate broad improvements under the same retained KV budget while preserving the practical efficiency of compressed decoding, including peak memory usage and long-context decode throughput.

Diffusion models have revolutionized generative tasks but incur high latency due to iterative denoising. While cache-based strategies accelerate inference by reusing intermediate features, they largely rely on static, sample-agnostic schedules. We argue that this rigidity overlooks two facts empirically validated in this paper: (i) generation difficulty varies across prompts, requiring adaptive resource allocation--complex inputs demand more computation while simpler ones require less; (ii) error sensitivity fluctuates across timesteps, where static policies may cache high-error steps or waste computation on low-error ones. We therefore propose OnlineCache, a dynamic caching framework that jointly learns when to cache and how to correct approximation errors. We leverage policy gradient to train a lightweight network for adaptive speed-quality trade-offs, and incorporate a learnable corrector to mitigate caching-induced errors. Both modules are jointly optimized under a bilevel optimization framework, with the policy targeting global generation quality and the corrector minimizing local errors. Our method automatically allocates computational resources across both samples and timesteps, improving overall generation quality. Extensive experiments demonstrate clear superiority. On FLUX.1-dev model, OnlineCache achieves nearly 3 speedup while preserving generation fidelity. On DiT and CogVideoX, it similarly delivers competitive acceleration without compromising quality; across all scenarios, it consistently outperforms existing cache-based acceleration baselines.

Genetic Programming Symbolic Regression (GPSR) generates mathematical expressions to model input-output relationships using an evolutionary process. A significant challenge in GPSR lies in the repeated evaluation of entire expressions or their sub-expression, which inflates computational runtime. To address this inefficiency, caching mechanisms have been employed to reduce redundant computations. However, prior studies predominantly employ a single caching strategy, offering limited insights into their comparative performance or memory-runtime trade-offs. In this paper, we present a comprehensive analysis of caching mechanisms for GPSR on synthetic and real-world datasets. We also include an empirical study of key-value usage frequencies under an infinitely large cache, offering insights into optimal cache sizing. Furthermore, we provide actionable guidelines for configuring caching strategies based on computational and memory constraints. Our findings indicate that complex caching mechanisms necessitate a minimum cache size to achieve computational time reductions. Conversely, lightweight caching strategies, such as Least Recently Used (LRU) and, notably, First-In-First-Out (FIFO), can significantly decrease computation time for fitness evaluations, which are a substantial component of the overall runtime.

Training-free acceleration makes diffusion-based multimodal large language models (dMLLMs) more deployable, but it may silently change generated content. We study this serving-time consistency problem on 300 real images, comparing Fast-dLLM outputs with the same model's unaccelerated outputs. Across the mild parallelism induced in our long-form setting (1.05--1.25 committed tokens per step), confidence-threshold tuning changes decoding behavior but not baseline agreement. State-refresh ablations and an image-swap intervention instead identify stale visual and generated-text states as contributors to drift. For the tested Fast-dLLM implementation, shortening the KV-cache refresh interval yields a monotonic speed--agreement frontier and near-exact agreement at a measured 1.3x speedup. The initial diagnosis also appears with dLLM-Cache and LaViDa, although dLLM-Cache recovers agreement only after both caches are tightened, which removes its speed advantage. Independent prompts and images reproduce the threshold-insensitivity and refresh recovery. A targeted audit finds genuine content substitution in half of 50 low-agreement pairs. In a separate blinded two-annotator evaluation, the pooled accelerated-minus-baseline factual-error difference is 0.00 (95% CI [-0.17,+0.17]); this sample detects no difference but does not establish factual equivalence. Finally, none of the tested adaptive or smoothed-refresh variants beats the fixed interval at matched compute. Our contribution is a paired diagnostic and an implementation-scoped consistency control, not an accuracy or safety guarantee.

Heterogeneous Large Language Model (LLM) systems increasingly rely on shared contexts, retrieved evidence, and multi-agent dialogue histories, yet their internal key-value (KV) caches remain model-specific and cannot be reused across architectures. Consequently, each model must repeatedly prefill or store caches for the same context, limiting the scalability of multi-model reasoning and long-context generation. We propose Mixture-of-Translators(MoT), a cache translation framework that maps context KV caches from a source LLM into the cache space of a target LLM. Unlike prior approaches that depend on a single projection path or global shared latent space, MoT uses multiple translator modules to capture diverse source--target mappings. To further reduce residual translation error, we introduce a Context Correction Loss that aligns the replayed target trajectory with the native target trajectory. We reveal two competing failure modes in cache translation: propagated translation shift from early injection and last-state shift from late injection. MoT addresses them through translator mixtures and target-side correction. Across homogeneous and heterogeneous translations among Qwen2.5, GPT-2, and OPT models, MoT preserves downstream QA performance, including Qwen2.5-7B-scale translation with 51.0% average closed-set QA accuracy and 0.43 average extractive QA F1. In practical case studies, MoT enables quality-preserving memory reuse for multi-agent reasoning and retains 96.3% of direct-context quality in long-context cache-augmented generation, demonstrating scalable KV cache reuse across heterogeneous LLMs.

LLM serving is increasingly accelerated by position-independent caching (PIC). Existing PIC methods, however, are built for full-attention models, where a token-indexed KV cache underlies its core operations: matching reusable token chunks, concatenating their KV entries, and selectively recomputing a few tokens to restore cross-chunk context. Hybrid LLMs break these primitives---they replace most attention layers with linear recurrences that expose only a fixed-size state, leaving no token-indexed KV to concatenate or to locally repair. This raises a natural question: can PIC benefit hybrid models, and what would it take? We present LinearKV, a training-free hybrid-PIC framework. Its key insight is a \emph{decoupled initialization}: each linear layer maps its $K$ matched local states to a single initial state, while full-attention layers concatenate their KV as before. LinearKV is therefore compatible with existing PIC methods, reusing their token selection and recomputation as-is. Under this framework, we find that a \emph{single cached state} suffices as the linear layer's initializer. The algebraically principled alternative---composing all $K$ cached states into the exact full-prefix state, as concurrent work HYPIC does---is unnecessary and, on some architectures, even harmful. We compare the two across three hybrid models and three PIC selectors. On the two GDN models the two tie, both recovering most of full quality (up to $92\%$); on the Mamba-2 model, exact composition instead collapses under every selector---under EPIC, for instance, it recovers only $46.6\%$ of full quality, versus $86.8\%$ for a single cached block initializer. A single state initializer is also cheaper, cutting time-to-first-token to $0.46\times$ full prefill versus a further $5$--$17\%$ overhead for exact composition; results hold across LongBench QA and RULER at 8K--32K.

Long visual context poses a challenge for vision-language models: performance degrades as the number of distractors grows, and processing all tokens at once is computationally infeasible under GPU memory constraints. We present ReToken, a single learnable embedding trained as an explicit retrieval target that selects a sparse set of query-relevant visual tokens from a pre-filled visual KV cache. Trained on only a small image-QA dataset, ReToken yields consistent gains across image and video benchmarks: on Visual Haystacks it improves Qwen3VL-8B by 13.4 points and InternVL3.5 by 12.4 points (>20% relative), and on LVBench it transfers zero-shot to long video for an 8.0-point gain with Qwen3VL-8B. Thanks to its lightweight design, both training and long-video inference fit on a single H100. Code is available at: https://github.com/avaxiao/ReToken

Stage-replay diagnostics reconstruct intermediate token prefixes and treat fresh-prefill continuation as continuation from the decoder state that originally reached the prefix. We audit that assumption at a whole reasoning-stage boundary in a Qwen2.5-derived system. A matched 200-item experiment compares retained live cache with one-shot prefill of identical integer tokens and places an exact replica on both sides. In BF16, replicas remain exact while the constructions differ on 166 suffixes and 20 correctness labels; the accuracy difference is only one point (paired 95% CI [-3.5, +5.5]). A fixed-prefix 2x2 holds all 200 token states constant while crossing construction and precision. The BF16 disagreements recur, whereas FP32 produces no decoded disagreement (95% Wilson upper bound 1.88%). A prospective bridge makes token-by-token incremental and retained live caches bit-exact on 12/12 rows; an all-200 saved-ledger audit reproduces every retained trajectory and comparison fingerprint. Bidirectional transplantation of all 48 key/value layers makes every tested divergent continuation follow its cache donor, both on a selected set at the primary checkpoint (24/24) and an outcome-blind replication at a later checkpoint (43/43). Exact-token replay can therefore be repeatable without preserving live-state fidelity. On the tested states, boundary K/V cache is a causally sufficient carrier of the divergent trajectory, while numerical precision moderates its behavioral expression.

Transformer depth is not used uniformly: lower and middle layers build semantic representations, while upper layers increasingly specialize them for prediction. We turn this division of labor into CoMem (Comprehension Memory), which writes each context chunk only through an intermediate layer, retrieves a fixed number of cached residual states, and recomputes the query-conditioned upper layers over the resulting pack. For a fixed retrieval budget, model-side read compute and memory are independent of stored-context length. We evaluate a continued-trained Qwen3-8B base LM under a unified chat-template-free protocol. The backbone is frozen; the flagship trains only a rank-32 self-distillation LoRA on plain PG19, and we report an adapter-free arm separately. CoMem reaches 97.05 on RULER and 38.27 on LoCoMo versus 34.59 for full-context KV-Direct; the dialogue-memory advantage survives conversation-cluster resampling and an independent judge. Results on additional long-context and long-document tasks expose both the benefits of bounded retrieval and its in-window compression tax. Controlled depth sweeps show that deeper caching lowers per-query recomputation but incurs a fidelity loss that self-distillation substantially repairs. In a separate adapter-free efficiency control on an NVIDIA H20 at 128k, CoMem uses 18.26 GB rather than 89.36 GB and achieves a 7.83x prefill speedup. These results show that long-context memory can be organized along the layer axis, not only the token axis.

Long-context retrieval and agentic workloads repeatedly reuse the same documents under changing instructions, histories, and document orders. Prefix caching cannot exploit this reuse, while position-independent caching (PIC) remains unreliable because independently compiled KV states lack the future context in which they will be consumed. Our diagnostics show that a learned boundary-conditioned baseline sharply reduces attention deviation near reusable-block boundaries but leaves interior and task-level residuals, motivating adaptation of the document representation itself. We present \emph{SemPIC}, which trains a LoRA-enabled Writer to compile native per-layer document KVs through behavioral distillation while retaining the pretrained decoder as an unchanged Reader. Adaptation is confined to offline cache construction, preserving the standard KV interface and cache-hit decoding path. We further introduce KV Gradient Checkpointing, which reduces peak training memory without severing gradients through cached KVs. Across three models and four tasks, SemPIC raises mean micro-F1 over KV Packet from 0.53 to 0.60, approaching Full Recompute at 0.62. Code: https://github.com/jn12-29/SemPIC

KV-cache quantization is validated today by offline benchmark averages; a deployed system cannot tell whether compression is damaging the request it is serving right now. We give it a provably sound runtime meter -- a "DTrace for KV quantization": a per-(layer, head, step) upper bound on the total variation between exact and compressed attention. The meter has two tiers: a deterministic band-norm-witness bound, sound for any cache-preserving black-box quantizer and for any query (adaptive-safe, worst-case Cauchy--Schwarz plus RoPE band-unitarity), and a tighter probabilistic certificate for a controlled subtractively-dithered INT8 quantizer under an explicit request-level failure budget (stated for non-adaptive queries; core theorems machine-checked in Lean 4). Three results. Observability: the meter enters SGLang through an env-guarded patch, and any scheme registered as one tensor function is measured in live serving. Repair: meter-driven gating -- risk-ranked where the witness is saturated, certified where it is informative -- empirically restores the quality floor at benchmark scale, e.g. raw-cast fp8 from 22.8 back to 79.7 on hard RULER tasks with the difference from uncompressed bounded at $[+0.0,+0.8]$ by a paired test. Analysis: aggressive schemes survive on cross-layer error cancellation, not per-step fidelity -- in a 28-layer sweep, no single layer's pollution alone loses anything (0/28) -- and the certified int8 cache serves $1.88\times$ more KV tokens at the same memory in SGLang. All artifacts, guards, and the Lean development are released at https://github.com/metask-ai/witcert-kv-certificates; every number regenerates from the shipped artifacts by one command.

Key-value (KV) cache management through compression and eviction strategies has emerged as an important research direction in recent years. Computational demands of large language models (LLMs) and their multi-modal variants during output generation can be partially alleviated by caching previous key and value calculations needed by subsequent scaled dot-product attention operations. However, this leads to another problem: the size of the resulting KV cache grows linearly with context length and quickly consumes all available GPU memory when either the prompt or the generated output are long. KV cache management periodically prunes entries from the cache thereby reducing its memory footprint while attempting to retain sufficient information for accurate generation. A by-product is faster inference speed. We propose a simple yet effective KV eviction scheme motivated by the insight that past tokens which can be well-predicted from more recent tokens are redundant and their associated keys and values can be removed from the cache. To score entries for eviction we run the model on the tokens in their original order, reusing the key and value representations already stored in the KV cache, and applying a counter-causal attention mask so that each position attends only to its future context. This is in-distribution, tied directly to the actual cache contents, and requires no additional training. To further reduce cost, we additionally propose a fast single-layer approximation that restricts the counter-causal pass to the last transformer layer, achieving a significant speedup per refresh cycle at marginal accuracy cost. We evaluate our strategy on various open-source LLMs and benchmark datasets showing competitive or improved performance over other state-of-the-art methods. Reference code is available at https://github.com/metacognitionai/counter_causal.

Large language models are increasingly deployed with persistent personalized context, such as accumulated memory profiles or long conversation histories, that is shared across a user's many requests. Production memory systems (e.g., Mem0, MemGPT, and Zep) retrieve a relevant subset of this memory and inject it into the prompt, forcing the serving engine to repeatedly prefill the same content. As the retrieval budget grows, time-to-first-token (TTFT) increases even though the underlying memory is reused across requests. We present InferScale, a GPU-native LLM memory system that replaces repeated prompt prefilling with reusable KV state. InferScale precomputes each memory fact's KV representation, stores it alongside a semantic embedding on the GPU, retrieves relevant facts at serving time, and injects their KV directly into vLLM's paged cache. To support dynamically assembled memories under rotary position embeddings, we introduce Chunked RoPE, which stores keys before rotation and applies their serving-time positions during injection. However, encoding memory facts independently omits the cross-fact context available during joint prefilling. We mitigate this with Context-Window Encoding, which encodes each memory fact together with a small window of preceding conversation context while caching only the target fact's KV. InferScale is implemented through vLLM's KV-connector interface, requiring neither engine modifications nor model fine-tuning. Across three open-weight models on LoCoMo, InferScale keeps TTFT nearly constant as the retrieval budget increases: at k=50 it reduces TTFT by 72-79% (3.6-4.8x), achieves 60.3% accuracy versus 63.3% for Mem0 without serving-time recomputation, and delivers 3.7-4.5x the throughput under concurrent load. Reusable KV state thus decouples memory-conditioned serving latency from retrieved-context size while preserving application quality.

Multi-head latent attention (MLA) is increasingly important for long-context LLM inference because compact latent states replace the growing key-value (KV) cache and reduce decoding memory traffic. Yet most capable open checkpoints use multi-head or grouped-query attention (MHA/GQA), so conversion is needed to obtain MLA's cache efficiency without retraining from scratch. Speculative decoding offers complementary acceleration, but its speedup depends on agreement between draft proposals and target verification. We find that direct MHA/GQA-to-MLA conversion can sharply reduce this agreement: low-rank factorization and RoPE handling introduce attention-function errors that may be tolerable for standalone generation but substantially lower draft-token acceptance. We therefore formulate MLA draft construction as functional reconstruction rather than cache compression. Our end-to-end (E2E) method optimizes each converted MLA attention module to reproduce the post-output-projection response of its original MHA/GQA counterpart on calibration hidden states. This converter-agnostic post-conversion procedure preserves the converted cache and inference graph and requires neither verifier logits nor verifier supervision. We evaluate 192 model-converter-backend-method-task configurations spanning four Llama/Qwen draft-target pairs, TransMLA and MHA2MLA, HF and vLLM, and four 200-prompt tasks. With a 0.5-percentage-point reporting tolerance, Functional Reconstruction materially improves acceptance in 37 of 64 matched task cells, leaves 26 practically unchanged, and materially decreases one. Code and evaluation artifacts are available at https://github.com/swyhahaha/FunctionalMLA.

Early ranking stages in recommendation systems precompute item embeddings and cache them in-model for scoring within strict latency constraints. Because this cache exists only at serving time, outside the training loop, training and serving use different item representations, a structural discrepancy that limits quality and adds operational fragility. We show that co-designing the training and serving paths removes this representation discrepancy at its source. We introduce the memory layer, an in-model key-value embedding cache co-trained with the model: the item tower writes embeddings during training and the model reads them at serving, one source of truth for item representations by construction. Always-on embeddings cover items not yet cached, so every item receives a prediction, and the design consolidates three separate trainer-to-predictor update paths into a single self-contained pipeline. Deployed in production on Instagram Reels, the memory layer raises prediction coverage from 96% to 100%, improves embedding freshness from $O(5\text{ min})$ to $O(20\text{ s})$, and narrows the training-serving Normalized Entropy (NE) gap by up to 86%, yielding over $2\times$ recall for the freshest content and a 5-6% cold start engagement lift. Because embeddings are produced during training, the system needs no separate bulk-evaluation or publish-time recomputation, cutting training-and-publish computational cost by 30% at neutral serving computational cost.

A language model with a bounded working memory must repeatedly decide which stored items to keep. Every deployed method decides the moment an item arrives, from the past (StreamingLLM, H2O) or from a guess about the future (SnapKV). We recast the choice as an estimation problem on a hidden signal, whether an item will be reused, placing existing methods on one axis, the commit lag $H$: online filters and learned predictors commit at $H=0$, while Belady's offline optimum sits where the whole future is known. The missing regime in between, fixed-lag smoothing, waits a bounded number of steps, observes which items a correct near-future prediction attended to, and only then commits. This measurement, demonstrated utility, turns Belady's unobservable future request into something we read off the model itself. We instantiate it as a training-free policy, RMM, a strict generalization of H2O that reduces to it exactly when the measurement is uniform. In controlled settings where reuse is endogenous and separated in time, demonstrated utility identifies used memory far better than accumulated attention, and a small bounded memory behaves like a much larger one. But on independent third-party benchmarks, run inside NVIDIA's KVPress harness against its own SnapKV, H2O, and StreamingLLM implementations, the advantage mostly disappears: RMM is on par with H2O for single-turn question answering and loses to both H2O and SnapKV in a streaming multi-turn setting. The cause is simple: on natural text the model is correct about most tokens, so weighting attention by correctness barely changes it, and demonstrated utility collapses onto accumulated attention unless reuse is sharp and endogenous, which standard benchmarks do not exercise. Our contribution is the framework and an honest map of when measuring beats accumulating, not a new state of the art.

Serving large language models at long context is bottlenecked by the key-value (KV) cache, which is read in full at every decode step. Attention keys are locally low-rank though globally high-rank: shared low-rank bases discard page-specific directions that a page's own compact basis retains. LOCKS gives every page its own spectral summary (resident, about a tenth the cache's size), reconstructs within-page logits, estimates each page's attention mass by log-sum-exp, and attends only the top pages; selection itself reads no candidate keys or values. Selecting on this summary alone stays within about a point of the full cache on long-document QA (LongBench-v1), tracks the read-every-key oracle on retrieval-dense RULER down to the smallest budgets, and shows its largest margins on long-form reasoning (AIME26, MATH-500), where baseline selectors collapse. At its shipped $2048$-token budget LOCKS matches FullKV aggregate quality at $100$K$+$ context while attending about $2\%$ of the tokens, and halves per-token decode latency ($2.0\times$ at $1$M tokens) against dense attention. LOCKS ships as a drop-in plugin for unmodified vLLM, with batched decode running in full CUDA graphs.

As the inference phase of Large Language Models (LLMs) requires handling long context windows, the Key-Value (KV) cache initially appears to address this challenge but eventually becomes a significant bottleneck as the context window continues to grow. Low-rank compression has recently been studied as an effective approach to reduce KV cache memory while maintaining model performance. However, only a few existing methods treat the Key and Value caches differently, despite their distinct roles. Moreover, these methods typically employ fixed attention-head grouping, which may not fully exploit the structural similarity among attention heads. In this paper, we propose an improved low-rank KV cache compression framework. For the Key cache, we dynamically group attention heads based on Centered Kernel Alignment (CKA) similarity and allocate the rank budget adaptively under a parameter budget. For the Value cache, we adopt the same approach as ReCalKV, refining the low-rank decomposition through offline calibration to improve reconstruction quality. Experimental results on three instruction-tuned LLMs show that our method reduces the number of Key cache parameters while maintaining competitive accuracy. We further observe that the proposed strategy is particularly effective for Multi-Head Attention (MHA) models, whereas it should be applied more conservatively to Grouped-Query Attention (GQA) models, especially in long-context settings.

High-resolution image and video diffusion models, including SD3, FLUX, and recent video diffusion transformers, have substantially improved generative quality but remain expensive at inference time because they repeatedly evaluate attention-heavy denoisers over many sampling steps. We address this inefficiency by exploiting redundancy in intermediate diffusion features rather than changing model weights or retraining. We identify four complementary redundancy sources in image and video generation: intra-frame, inter-frame, motion, and denoising-step redundancy. Based on this analysis, we propose OmniCache, a unified hierarchical caching framework that performs multidimensional feature reuse through Token Cache, Frame Cache, Block Cache, and Layered Cache. Unlike token-merging baselines that average matched features, OmniCache uses similarity matching to select cacheable features, skips redundant computation, and restores positionally consistent cached activations, preserving feature order and spatial-temporal structure. The resulting framework reuses spatial features in temporal layers and temporal features in spatial layers, while Layered Cache captures cross-step redundancy at the model-layer level. Across SD3, SVD-XT, and Latte, OmniCache reduces inference latency by up to 35%, 25%, and 28%, respectively, while maintaining visual fidelity and motion coherence in a training-free setting.

Test-time search lets small video diffusion models rival larger ones, but costs 2-10x more. All candidates are fully denoised, although most are discarded. Training-free caching makes each rollout 2-3x faster at near-lossless quality. Composition is safe only if lossy caching preserves verifier rankings. We present the first study of whether caching corrupts candidate ranking in video test-time search. On Wan2.1-T2V-1.3B with an adaptive caching wrapper (~2x per-candidate speedup), ImageReward scores seed-matched cached and full rollouts. Median per-prompt Spearman rank correlation is 0.905, with 72% top-1 agreement on the VBench suite. VBench-2.0 replicates this result on a harder suite. Recomputing the cached winner at full compute retains 90-94% of the full-search gain. Errors cluster among near-tied candidates, making corruption self-limiting. This finding leads to CachedSearch. It explores every candidate with aggressive caching, then re-generates only the winner at full compute. At N=8, it captures 94.7% of best-of-N's gain at 63% of the cost. Capture rises with width. At matched budget, it searches twice as wide for 38% more gain. The result holds from 1.3B-14B across six models and four families: Wan, LTX, CogVideoX, and Hunyuan. Wan2.1-14B matches the 1.3B model's fidelity. Mid-trajectory pruning multiplies the exploration saving to 3.11x at 88.6% capture. Ports to other model families require recalibrating a single parameter, showing that fidelity tracks architecture rather than parameter count. CachedSearch is training-free, verifier-agnostic, and orthogonal to the search algorithm, making it a plug-in multiplier for test-time scaling.

With the rapid adoption of long-context large language models (LLMs), the continuously growing KV cache during decoding has become the critical memory bottleneck. To tackle this challenge, we propose HiKV, a novel algorithm-hardware co-design that exploits KV cache redundancy through hierarchical importance awareness. Algorithmically, HiKV compresses the KV cache at two granularities: Stage I evicts unimportant tokens within a fixed budget, and Stage II further loads only the significant elements of each retained token, reaching compression ratios unattainable at a single granularity. Architecturally, we develop a dedicated accelerator centered on a reconfigurable importance sorter that switches between the distinct sorting datapaths each stage requires, unifying the two-stage acceleration in one circuit with minimal overhead. Evaluated on representative LLMs, HiKV achieves up to 7.95x speedup and 90% energy reduction in the attention computation over the vanilla KV cache baseline within negligible 1% accuracy loss. Under iso-accuracy constraints, HiKV outperforms state-of-the-art importance-based methods by achieving an additional 1.82~4.87x reduction in external memory accesses. These benefits are enabled by specialized hardware components that add only 8% to the system area.

Deterministic KV-cache eviction keeps the top-$k$ tokens under an importance score and deletes the rest. We prove that this design cannot know what it destroyed: evicted values can be altered so that everything the serving system retains is unchanged while the true attention-output error grows arbitrarily, so no serving-time estimator of that error is consistent. Randomized eviction restores identifiability. With a Poisson-sampled tail at known inclusion probabilities, one logit offset performs the Hájek correction inside the softmax, and a survey-sampling variance estimator over the retained set becomes a per-step error certificate with 0.97 empirical coverage at no accuracy cost. On real workloads, seven pre-registered claims locate the certificate's value precisely. Prediction goes to output confidence: question-aware eviction at 25--50% budgets is nearly free, output log-probability predicts failure better than any cache-side signal, and certificate-gated budget escalation adds nothing. Attribution stays with the certificate: it separates cache-induced from inherent failures (AUC 0.65--0.75, against 0.47--0.54 for output confidence) and schedules recomputation better than random or confidence gating. Randomization buys attribution, not prediction.

Diffusion policies achieve strong visuomotor control by iteratively denoising action chunks, but repeated denoising makes real-time deployment computationally demanding. Cache-based methods reduce inference cost by reusing intermediate activations, but existing training-free schedules typically allocate computation uniformly across blocks, ignoring heterogeneous redundancy across blocks and leading to a suboptimal performance-efficiency trade-off. To bridge this gap, we introduce Evolving Cache Schedules (EVO), a training-free acceleration framework that globally schedules cache refreshes via evolutionary search. EVO represents each candidate as a complete schedule over the block-timestep lattice. Thus, redundant transformer computations during iterative denoising can be skipped through cache reuse while preserving closed-loop rollout performance. To make the search practical, EVO introduces redundancy-aware initialization, which seeds the population with promising schedules, and target-conditioned early stopping, which verifies and terminates once a desired performance target is reached. The offline-optimized schedule can be directly plugged into pretrained diffusion policies without retraining. Extensive manipulation benchmarks show that EVO preserves near-full performance while substantially reducing computation, achieving up to 8.05x action-generation speedup and reducing FLOPs from 15.77G to as low as 1.96G. Source code is available at https://github.com/pillom/EVO.

Autoregressive (AR) video diffusion models have become a promising paradigm for long and streaming video synthesis, but the continuously growing Key-Value (KV) cache makes attention the dominant inference cost, especially at high resolution where each frame contributes many tokens. Existing remedies either evict the cache with coarse heuristics that cause inter-frame flickering, or require model re-training. We propose HeadCast, a training-free, plug-and-play acceleration framework built on the observation that a pre-trained AR model's attention heads exhibit stable, heterogeneous behaviors. After a short warm-up, HeadCast performs a one-time classification at the maximum-noise step that sorts every head into one of four archetypes: Sink, Dummy, Spatial, and Global, and restructures the monolithic KV cache into head-specific pathways. Crucially, it retains the Global heads that preserve the long-range temporal consistency aggressive eviction destroys. Because the Spatial pathway operates on a fixed-size grid, its savings grow with resolution: across state-of-the-art AR models, HeadCast accelerates inference by up to 1.62x at 720P and 1.95x at 1080P, while keeping VBench quality on par with full attention and largely flicker-free. Code is available at https://github.com/sjlgaga/HeadCast .