Embedded devices typically lack the resources of GPU-equipped machines, and existing inference methods suffer from either high computational overhead (patch-based) or accuracy loss (approximation-based). We propose GaLe, a memory-efficient technique that enables the deployment of pretrained networks on constrained devices without retraining. GaLe partitions feature maps into two components: a local exact (Le) representation that preserves fine details and a global approximate (Ga) representation that retains long-range dependencies. Unlike standard tiling, GaLe supports global operations and attention mechanisms found in hybrid CNN-transformer models. Validated on ImageNet, our method matches exact-inference performance while achieving up to 65% speedup and 90% RAM reduction on a Cortex-M33 compared to patch-based inference. We further demonstrate GaLe's versatility across classification, detection, and generation tasks, highlighting its potential as a foundation for resource-efficient architecture design.
Diffusion models achieve high sample quality but remain expensive at inference time because sampling requires many sequential neural function evaluations (NFEs). Existing acceleration methods either use fixed step-skipping schedules, adapt step sizes based on local numerical error, or require additional training. We introduce GeoSPRINT (Geometric Step Pruning for Inference in Trajectories), a training-free framework for constructing non-uniform sampling schedules from the geometry of denoising trajectories. GeoSPRINT detects geometrically redundant steps using a hyperplanarity test in latent space, implemented efficiently via QR factorization, and converts the resulting redundancy profile into a sampling schedule that allocates more steps to high-curvature regions of the trajectory. In addition, we introduce the trajectory projection score $α_{\mathrm{traj}}$, a residual-variance metric that quantifies trajectory straightness and serves as a model-free diagnostic for rectified flow quality. Across CIFAR-10 ($32{\times}32$), LSUN Church ($256{\times}256$), and Stable Diffusion v1.5 ($512{\times}512$ latent), GeoSPRINT consistently improves over uniform DDIM (Denoising Diffusion Implicit Models) schedules at matched NFE budgets. On CIFAR-10, GeoSPRINT improves FID (Fréchet Inception Distance) by 0.7-1.1 over DDIM across 49-89 NFEs and surpasses DPM-Solver++ at NFE${\geq}30$ despite using a first-order DDIM solver. On LSUN Church, it reduces FID from 1.48 to 1.26 at 52 steps, and on Stable Diffusion v1.5 it achieves up to 1.93 FID improvement over DDIM. These results show that trajectory geometry provides a useful global signal for allocating inference steps and that schedule quality can substantially improve diffusion sampling efficiency without retraining.
Few-step diffusion models substantially compress temporal computation, making the spatial cost of each model evaluation an increasingly dominant source of inference latency. Progressive-resolution inference reduces this cost by performing early denoising at low resolution and reserving high-resolution computation for refinement. However, existing methods typically lift intermediate latents directly and rely on subsequent steps to absorb the induced distribution mismatch. In the few-step regime, the limited recovery budget leaves these errors as visible artifacts, constraining how late the transition can occur and, consequently, how efficiently it can be performed. We introduce SelfLift, a self-recovering progressive-resolution framework that derives both transition-repair signals and trajectory-aligned supervision from the generative model itself. SelfLift-zero proposes a training-free Artifact-Aware Consistency Lift, using disagreement between direct latent lifting and pixel-VAE re-encoding as both a localized artifact-risk signal and a model-native correction direction. It enables reliable late transitions without external super-resolution, extra denoiser evaluations, or sampling-schedule modifications. Building on this robust transition, SelfLift-rich performs On-Policy Self Recovery on student-visited states, transferring dense high-resolution guidance from an internal self-teacher while remaining aligned with the altered progressive-resolution dynamics. Across FLUX.2-Klein and Z-Image-Turbo, SelfLift reduces end-to-end latency by 41.5% and 44.1%, respectively. Combined with timestep distillation, it delivers overall speedups of 29.61x and 19.21x over the corresponding 50-step models while preserving competitive generation quality, establishing a stronger speed-quality frontier for few-step diffusion.
Alessandro Zirilli, Davide Marincione, Evgenios M. Kornaropoulos +2cs.CR cs.AI
Fully homomorphic encryption (FHE) allows a server to run a language model directly on encrypted user prompts, but current approaches remain prohibitively slow. Ciphertexts natively support only addition, multiplication, and rotation, and multiplications may be composed only to a bounded depth before a costly bootstrapping operation is needed to continue. Every nonlinearity must therefore be approximated by an iterative method, and each iteration uses multiplications. A higher iteration count buys precision but exhausts the available depth faster and triggers more bootstraps, which dominate latency. Existing approaches fix the iteration counts uniformly across the model rather than tailoring them to each site's error tolerance. We introduce Homomorphic Encryption-Aware Training (HEAT), a fine-tuning method that makes the per-nonlinearity iteration counts learnable, enabling them and the model weights to co-adapt during training. HEAT optimizes iterations with respect to the task objective, allowing the model to adapt to approximation errors encountered during inference without architectural changes or retraining from scratch. On encrypted GPT-2 decoding, HEAT reduces iterations by $3.1\times$, bootstraps by $1.6\times$, and end-to-end latency by $1.4\times$, while improving decode agreement over the calibrated baseline.
With the rapid and continuous growth in the incorporation of machine learning models based on the Transformer architecture, capable deployment is in high demand. In this context, capable deployment refers to operational performance aspects, e.g., throughput and latency, as well as efficiency aspects, e.g., energy consumption. When it comes to the task of inference using such models, purpose-built hardware accelerators provide a lucrative alternative to common deployment choices, such as Central Processing Units (CPUs) and Graphics Processing Units (GPUs). The Field Programmable Gate Array (FPGA) platforms category is an example of such alternative accelerators, promising implementation flexibility, energy efficiency, improved latency and suitability for on-site deployment. We investigate the most recent advances, trends, and design choices for Transformer inference on FPGA platforms. We perform a systematic literature review, extracting and delving into preferred techniques for implementation and optimisation. This study and the provided taxonomy of topics could act as a guide for researchers from the academia and industry alike.
Mixture-of-Experts (MoE) architectures scale Large Language Model (LLM) capacity efficiently by activating a sparse subset of experts per token. However, modern MoE inference remains heavily constrained by the rigid, whole-expert abstraction. Existing frameworks manage, schedule, or prune experts as atomic execution units, which fixes the optimization boundary too early and leaves fine-grained intra-expert computational redundancy underexplored. In this work, we present PCoMoE, a path-compositional execution framework that shifts MoE inference from coarse-grained expert selection to fine-grained path composition. PCoMoE incorporates a path-level formulation of expert computation, a compatibility-aware layer-wise pruning strategy to suppress low-value path combinations, and a hardware-friendly execution engine to exploit reusable sub-expert structures under strictly bounded overheads. Experimental results demonstrate that PCoMoE achieves up to a 1.31x end-to-end inference speedup while enhancing model accuracy by 10%. The code is available at https://github.com/gzyyy0/PCoMoE
Jungseob Lee, Seongtae Hong, Dongyub Jude Lee +4cs.AI cs.CL cs.CV
Speculative decoding accelerates generation without changing its output, yet on vision-language models (VLMs) it has been caught in a self-defeating cycle. The drafter stays autoregressive, so it must stay small. A small drafter cannot afford the image at every step, so vision is compressed, pruned, or hidden. A drafter cut off from the image is then least reliable exactly where the image makes text predictable. We present GLANCE, the first one-pass block drafter that is lossless on an unmodified VLM target, and it breaks the cycle at both ends. A block-diffusion head reads the target's already-fused vision-language state, so vision costs the drafter nothing, and fills a whole block in one forward pass, so depth costs no sequential steps. A wide candidate tree is verified in one target pass, and every audited prompt reproduces greedy decoding exactly. Grounded workloads reward this most, entering a verbatim-copy regime whose long runs cost an autoregressive drafter a pass for every token and a block drafter one in total. Under one engine and one round budget, GLANCE decodes up to 2.93x faster than autoregression, from one draft pass a round where the production EAGLE3-VL head takes eight, and accepts 2.7x longer blocks than an EAGLE-3 head trained on the same corpus. One law organizes these results. Accepted length is set by the target's next-token entropy, with a fitted slope that steepens with grounding across all five tasks. The law transfers across targets and modalities and names its own boundary, since free-running text still favors a chain. Our code is available at https://github.com/js-lee-AI/GLANCE.
Large language models (LLMs) have demonstrated strong capabilities in handling long-context tasks, but processing such long contexts remains challenging due to the substantial memory requirements and inference latency. In this work, we discover that certain attention heads exhibit sequential consistency in their attention patterns, which can be persistently identified using a coefficient-of-variation-based algorithm. Inspired by this observation, we propose CateKV, a hybrid KV cache method that retains only critical token information for consistent heads, thereby reducing KV cache size and computational overhead, while preserving the majority of KV pairs in adaptive heads to ensure high accuracy. We show the unique characteristics of our algorithm and its extension with existing acceleration methods. Comprehensive evaluations on long-context benchmarks show that, while maintaining accuracy comparable to full attention, CateKV reduces memory usage by up to $2.72\times$ and accelerates decoding by $2.18\times$ in single-sample inputs, and boosts throughput by $3.96\times$ in batch scenarios.
Long-context LLM applications such as document summarization and multi-turn agents require generation from prefixes spanning tens of thousands of tokens, making decoding latency a major bottleneck. Speculative decoding (SD) reduces latency without changing model outputs, but its speedup depends on both accepted draft tokens and draft-step latency: Lightweight drafts are fast but lack the capacity to capture long-range dependencies, whereas strong independent drafts recover acceptance but incur growing KV-access cost at long prefixes. We introduce memory-augmented drafting for long-context SD, equipping a strong independent draft with compressed draft-side KV memory: A lightweight adaptor constructs and incrementally updates this memory to retain distant information and exact recent context. The target verifier retains its full KV cache and applies the standard accept/reject rule, preserving SD's lossless guarantee. Experiments on Llama~3.1-8B and 70B targets at prefix lengths up to 32K show that our method reduces draft-side memory by over 70%. It achieves speedups of up to 2.08x and 3.33x , respectively, over autoregressive decoding.
Speculative decoding accelerates large language model inference by using a draft model to generate candidate tokens, which are verified by the target model in a single forward pass. Verification proceeds sequentially and discards every position from the first rejection onward, yet existing draft training relies on token-level imitation of the target with a fixed per-position weighting that reflects neither property. We introduce Verification-Aware Training (VAT), a plug-in framework that simulates verification at every training step and turns the resulting accept and reject patterns into supervision. VAT consists of two components: (i) a verification head, a lightweight jointly trained binary classifier that supervises the draft model on whether each position survives sequential verification; (ii) verification-adaptive weighting, which replaces the fixed weighting schedule by keeping full weight up to each sample's first rejection point and re-anchoring the decay to start there. VAT modifies only the training objective, so it can be layered on top of existing methods without changing the draft architecture, the target model, or the inference procedure. Applied to EAGLE-3 and DFlash on Qwen3-4B, Qwen3-8B, and LLaMA-3.1-8B, VAT improves average acceptance length by up to 11.4% and wall-clock speedup by up to 8.7%, with consistent gains across math, code, and chat benchmarks. Code will be available at https://github.com/naver-ai/vat
Speculative decoding accelerates autoregressive language model inference by having a lightweight draft model propose multiple candidate tokens, which are then verified in parallel by a larger target model. However, after the first rejection, standard prefix-based verification discards the remaining draft suffix, so the computation spent generating and verifying those positions does not contribute to decoding progress. Focusing on DFlash, we show that rejected positions in a rejected suffix may still align with the target continuation, indicating that the draft model can retain useful semantic and structural information despite local token-level errors. Motivated by this observation and inspired by conditional diffusion, we introduce~\textbf{ReTrace}, a rejected-trajectory conditioning method that conditions each draft block on the rejected suffix from the previous round rather than generating it from fresh mask placeholders alone. ReTrace retains the hidden representations of the rejected suffixes, aligns them with the next draft block, refines them using target-aware correction signals from the same verification pass, and admits them into the drafter's input embeddings through gated residual fusion. Because rejected tokens are never committed and target-side verification remains unchanged, ReTrace preserves the lossless property of speculative decoding without requiring an additional model forward pass. Experiments with Qwen3 models across mathematical reasoning, code generation, and open-ended dialogue demonstrate that ReTrace consistently improves average acceptance length and end-to-end decoding speed over its DFlash backbone. By introducing cross-round conditioning without modifying within-round proposal generation, ReTrace is largely orthogonal to existing drafting improvements and might be combined with them for further gains.
Diffusion-based visual generative models deliver strong image and video synthesis quality but incur high inference costs because sequential samplers repeatedly evaluate large networks. Caching-based methods reduce inference latency by reusing intermediate computations across adjacent timesteps. However, existing cache controllers rely primarily on local temporal variation and overlook the trajectory-level consequences of cache reuse. We introduce Error-Propagation-Aware Cache (EpaCache), a training-free caching policy that adaptively allocates the reuse budget on timesteps with lower downstream impact. Experiments on image and video synthesis models demonstrate that EpaCache consistently improves the latency--fidelity trade-off over existing caching methods. On FLUX.1-dev, EpaCache outperforms the prior state-of-the-art caching method in both latency and fidelity, reducing inference time from $11.7$ s to $11.3$ s while improving PSNR from $21.4$ to $22.8$. On HunyuanVideo, EpaCache achieves a $2.63\times$ speedup over uncached inference and improves SSIM from $0.891$ to $0.905$ over the prior state-of-the-art method at matched latency.
Vision-Language-Action (VLA) models, built upon Vision-Language Models (VLMs), have significantly enhanced robotic capabilities by leveraging internet-scale knowledge and multimodal reasoning. However, the intensive computational overhead of VLAs constrains on-device deployment, hindering real-time responses to environmental changes. While various acceleration techniques have been proposed, they often rely on fine-tuning or access to training datasets, which are frequently unavailable due to privacy and proprietary concerns. Moreover, although flow-matching-based VLAs have emerged as efficient alternatives to standard diffusion models, current acceleration efforts largely target VLM inference costs, failing to address the iterative ODE solving process inherent in flow matching inference. To address these limitations, we propose AdaVLA, an online, training-free adaptive framework for fast yet accurate flow-matching-based Vision-Language-Action models. We introduce a novel metric derived from the flow matching trajectory curvature to quantify action generation confidence during inference. This metric enables the dynamic reduction of inference steps and the adaptive adjustment of MLP pruning ratios through an efficiently computed importance evaluation, requiring no access to training data. Experimental results on the LIBERO benchmark using a Jetson AGX Orin device demonstrate that our method achieves $1.87\times$ and $2.24\times$ speedups for $π_{0.5}$ and X-VLA, respectively, with negligible degradation in success rates. Furthermore, we validate the robustness of our approach on real-world robotic tasks using SmolVLA.
Dynamic sparse attention can reduce the quadratic cost of long-context prefilling without changing model weights. MInference assigns each attention head one pattern offline and estimates that pattern's sparse indices for every prompt. This design is efficient, but it assumes that a head's preferred pattern and sparsity budget remain suitable across inputs. We introduce RouteSparse, which routes each head and prompt segment among a small library of GPU-efficient sparse patterns. A low-cost probe estimates pattern utility and uncertainty; a latency-aware router then selects a pattern and budget, while uncertain cases fall back to a denser mask. We formulate routing as constrained risk minimization, derive an attention-output error certificate from omitted probability mass, and evaluate the method on long-context retrieval, question answering, summarization, and language modeling. On Llama 3.1-8B-Instruct with 128K-token prompts, RouteSparse achieves $6.5\times$ dense prefill speed with a 0.2-point RULER drop relative to dense attention, compared with $7.3\times$ speed and a 1.6-point drop for fixed per-head routing. Ablations confirm that input-conditional routing, hardware profiling, and selective dense fallback each contribute to the quality--latency tradeoff.
With the growing demand for processing multiple image sequences in real-world applications, various visual token pruning methods have emerged to mitigate the computational and context length constraints faced by Large Vision Language Models (LVLMs). However, most existing pruning approaches rely on static strategies that struggle to adapt across different architectural LVLMs and multi-image scenarios, and are additionally constrained by their dependence on attention computations that are incompatible with efficient techniques like FlashAttention. To address these limitations, we propose a training-free, Adaptive Visual Token Pruning (AVTP) framework, applicable to diverse LVLM architectures. We strategically determine pruning layers based on empirical analysis of visual attention distributions across various LVLMs, and implement adaptive pruning ratios in multi-image contexts where images of higher importance retain proportionally more tokens. We conduct extensive experiments across different LVLMs to demonstrate the effectiveness and robustness of AVTP. Specifically, Qwen3VL-8B achieves 2 times inference speedup while maintaining 96.1\% of its original accuracy on multiple multi-image benchmarks, InternVL3.5-8B retains 94.1\% accuracy, and LLaVA-OV-7B even exceeds its original baseline performance. Our code is available at \href{https://github.com/zry13/AVTP}{this link}.
Joint text-to-video-audio generation produces synchronized visual and acoustic content, but the long sampling trajectories and heterogeneous multimodal computation of large models make inference prohibitively expensive. We present TurboT2VA, a distillation and inference framework for accelerating a 19B-parameter joint video-audio model. Large-scale T2VA distillation is challenged by modality-imbalanced optimization, the difficulty of continuous-time consistency training at scale, and the quality--diversity trade-off. TurboT2VA addresses these issues with per-modality normalization and a progressive curriculum comprising discrete consistency warm-up, continuous consistency refinement, and joint consistency--distribution matching. The curriculum first establishes a stable, diverse generation trajectory and only then introduces distribution-level refinement. On LTX-2, four-step distillation reduces generator latency from 50.52s to 2.51s at the standard evaluation resolution of 512$\times$768, achieving a 20.1$\times$ speedup while maintaining strong visual quality, audio fidelity, diversity, and video-audio synchronization. We further develop an architecture-aware inference stack that combines guarded W8A8 and fused operators, padded-text compaction, and modality-aware sparse attention while preserving dense cross-modal and text-conditioning paths. Under the high-resolution deployment setting at 1024$\times$1792, the complete stack reduces generator latency from 318.74s to 5.83s on one NVIDIA H20, achieving a 54.67$\times$ generator-only speedup. Inference code and generation demos are available at https://github.com/thu-ml/TurboDiffusion/tree/main/turbot2va.
Large language model (LLM)-based agent applications often incur high response time. Speculative decoding is a promising solution to improve the inference efficiency of LLM agents without impacting generation quality. However, state-of-the-art speculative decoding algorithms exhibit substantial speed degradation under large batch sizes, limiting their effectiveness to deploy in real-world agent applications. In this work, we first present a systematic analysis of speculative decoding for LLM agents and identify two dominant factors of speedup degradation: high rejection rate of speculative tokens, and under-utilization of dynamic token budgets.B ased on these observations, we propose AgentSpec, a speculative decoding algorithm that addresses the limitations of existing methods for LLM agents. AgentSpec incorporates structure-isolated drafting that constrains speculation to semantically coherent segments of the agent workflow, reducing the drafts of irrelevant semantic paths and achieving an extremely low rejection rate. Moreover, AgentSpec adopts redundancy-aware budget allocation that exploits agent-level information to better utilize the dynamically-free token budget during the agent inference. We implement and evaluate AgentSpec on five different workloads and four different models from four different LLM families in vLLM. Our results demonstrate the superiority of AgentSpec over state-of-the-arts.
Diffusion Transformers achieve high-fidelity image and video generation, but their iterative sampling remains expensive, for each denoising step requires large matrix operations. Existing cache-based acceleration reduces redundant computation yet increases the VRAM footprint by storing intermediate states, which can directly constrain inference batch size. In this work, we propose a training-free acceleration method that performs stepwise forecasting for DiT sampling using a Barycentric Extrapolator. By leveraging barycentric extrapolation, our predictor is numerically stable and alleviates oscillatory artifacts analogous to the Runge phenomenon during forward forecasting. Across extensive experiments on both image and video generation, our approach provides a favorable trade-off between memory usage and perceptual quality, while delivering up to 3.30x end-to-end sampling speedup compared with baseline DiT inference.
Chengjie Lu, Tianchi Deng, Zhengqi He +2cs.LG cs.AI
Diffusion Transformers (DiTs) have shown strong performance in high-fidelity image generation, but their sampling process remains computationally intensive due to full model execution at every timestep. While cache-based acceleration has been explored to mitigate inference cost, naive reuse schemes suffer from low accuracy over long intervals, and Taylor-series-based extrapolation methods often face instability caused by Runge oscillations. In this paper, we propose ChebBooster, a training-free extrapolation framework based on Chebyshev polynomial theory that achieves stable and efficient acceleration for DiTs. Specifically, we adopt the Barycentric formulation to evaluate Chebyshev approximants with high numerical stability and minimal overhead, and further decouple the extrapolation into an offline weight precomputation phase and a lightweight online application stage. Extensive experiments across three representative DiT-based models, including DiT-XL/2, PixArt-$Σ$, and FLUX.1-dev, demonstrate that ChebBooster achieves consistent improvements in visual quality and inference efficiency, reaching up to $3.68\times$ latency speedup and $5.12\times$ FLOPs reduction, outperforming existing training-free baselines under diverse generation tasks and resolutions.
Mixture-of-Experts (MoE) models scale capacity for strong quality while keeping per-token compute bounded through sparse expert activation. Yet low-latency MoE serving is increasingly challenging, because it spans two inference phases with fundamentally different bottlenecks: prefill is dominated by token-wise expert computation, whereas decode is constrained by memory traffic from the batch-wise activated expert set. However, existing training-free acceleration methods optimize only a single resource proxy, either the experts each token executes or the experts a batch activates, and either discard the excluded experts' contribution or leave it only implicitly approximated. In this paper, we propose ExFold, a unified training-free expert-folding framework for jointly accelerating MoE prefill and decode. ExFold casts both prefill and decode as one budgeted output-approximation problem: execute only a phase-specific constrained expert set while projecting the contribution of budget-excluded experts onto retained experts using calibrated scalar projectors. Motivated by the observation that many expert outputs are directionally aligned but differ in magnitude, ExFold calibrates a pairwise scalar-projector matrix on unlabeled data and uses it at inference time to fold excluded expert contributions into retained experts. Under this view, prefill acceleration becomes token-level Top-K folding, and decode acceleration becomes batch-level expert-pool folding. The two phases differ only in how retained experts are selected, while excluded contributions are recovered by one shared folding mechanism. We implement ExFold as a plug-and-play plugin in vLLM, with a lightweight expert-folding CUDA kernel, delivering up to 1.41x TTFT and 2.45x TPOT speedups while retaining about 99% of the original average quality.
Farhana Amin, Sabiha Afroz, Dimitrios S. Nikolopouloscs.AI
Diffusion language models can generate many tokens in parallel, but they still require repeated denoising steps during inference. This makes generation costly, especially when the model continues to recompute tokens that are already stable. To address these limitations, we propose CAI-DLLM, a training-free inference method that uses first-step confidence to guide denoising and reduce inference time. Specifically, CAI-DLLM commits easy tokens earlier, allocates more denoising steps to harder tokens, and adjusts decoding schedules across output blocks. As it relies only on first-step confidence signals, it does not require retraining, extra predictors, or weight updates. We evaluate CAI-DLLM on LLaDA-8B-Instruct and Dream-7B-Instruct across math, code, reasoning, commonsense, and long-context tasks. CAI-DLLM achieves up to 18.2x wall clock inference speedup on LLaDA GSM8K while improving accuracy from 76.27% to 77.41%, and up to 13.1x speedup on Dream HumanEval while achieving higher pass@1 than no-cache inference, 48.17% compared with 46.95%. On harder reasoning tasks, speedups reach 44.8x, with a largest accuracy drop of 4.4 points, while energy consumption is reduced by up to 95.3%.
Large language models have become increasingly capable agents for low-level code and kernel optimization, but isolated kernel benchmarks provide only a proxy for the deployment behavior that matters in language-model inference. We identify a benchmark-to-deployment gap: candidate kernels that appear correct and fast in standalone harnesses can exhibit different performance, safety, or phase behavior after integration into a real inference workload. We introduce LLM4LLM, a deployment-aware closed-loop optimization framework that starts from a target inference script, extracts phase-aware optimization tasks, searches with an experience-guided episodic agent, and accepts patches through in-model validation. Across ten language-model inference workloads on A100 and H100 GPUs, LLM4LLM improves end-to-end latency for every evaluated model, achieving 3.91$\times$/6.98$\times$ geometric-mean speedups on A100/H100; as supporting kernel-level evidence, it also attains up to 2.745$\times$ GeoMean speedup on KernelBench Level 2.
Speculative decoding accelerates language-model inference by drafting future tokens that the target model verifies in parallel. A diffusion-style block head such as DFlash is an attractive drafter, predicting an entire block of future tokens in one forward pass. However, it is trained on per-position marginals rather than the joint block distribution, so the tokens it emits are individually plausible yet jointly incoherent. We introduce LiLiCorr, a Lightweight Likelihood-based model that Correlates the per-position marginal distributions a drafter already produces. It keeps the top-k tokens at each position as candidates and processes them jointly, producing for each an in and an out vector. A pair of adjacent candidates matches when the earlier one's out vector has high cosine similarity with the later one's in vector. These matches capture the block's joint structure without ever materializing the full joint distribution. One lightweight network pass produces all the vectors, and the pairwise scores are then computed in parallel as batched matrix operations, leaving only a cheap greedy walk sequential. We further co-train the drafter with LiLiCorr, so it learns to propose candidates that correlate into longer accepted sequences. Over the vanilla DFlash drafter, LiLiCorr raises acceptance length on every benchmark by 9 to 19%, while its scoring head accounts for about 2.8% of the per-block latency. Against DFlash and two concurrent methods that also restore coherence at draft time, LiLiCorr delivers the highest throughput in 70 of 72 settings: nine benchmarks at two target sizes under greedy and temperature-one decoding, and a throughput sweep over six concurrencies, two input lengths and three entropy tiers, with all systems equally optimized on a common serving stack. Extending LiLiCorr to inputs an order of magnitude longer than it was trained on preserves that lead.
Long-context modeling is a pivotal capability for Large Language Models, yet the quadratic complexity of attention remains a critical bottleneck, particularly during the compute-intensive prefilling phase. Our previous work, FlashPrefill, mitigates this cost through instantaneous pattern discovery and max-based dynamic thresholding; however, it remains an algorithmic prototype that is still distant from production deployment. In this paper, we present FlashPrefill V2, which evolves FlashPrefill from a prototype toward practical long-context serving along three dimensions. First, we introduce a mean correction term that effectively suppresses the approximation error, keeping performance degradation manageable even at extreme sparsity levels. Second, we redesign the sparse attention operator with PackGQA memory access, warp specialization, and pingpong pipelining, fully aligning with the latest FlashAttention-3/4 implementations and supporting FP8 inference to meet practical quantization requirements. Third, FlashPrefill V2 natively supports paged KV cache and continuous batching, allowing integration as an attention backend in modern inference frameworks such as SGLang. Extensive evaluations on NVIDIA H20 GPUs---among the most widely deployed inference accelerators---demonstrate that FlashPrefill V2 delivers up to 47.26x and 27.19x speedups over FlashAttention-2 at 128K context length under FP8 and BF16 precision, respectively, and, in FP8, still achieves a 30.49x speedup against an FA3/4-aligned dense baseline.
Venkat R. Dasari, Jakob A. Adams, Vinod K. Mishra +1cs.AI
Artificial intelligence (AI) models have demonstrated remarkable capabilities across various domains, yet their widespread deployment is impeded by significant computational costs, particularly on resource-constrained devices. This paper explores the theoretical underpinnings of various AI model optimization techniques, algorithms, and abstractions, discussing their potential to reduce computational complexity, memory footprint, latency, and power consumption. Furthermore, we propose a comprehensive hardware (HW) and model-agnostic generalized optimization architecture that integrates these techniques for improved efficiency. Our study underscores the critical role of such a generalized optimization system in preparing model deployment over resource-constrained heterogeneous hardware in a tactical environment. As a concrete demonstration, we show that GOE-compressed language models deploy and run on a GPU-less edge CPU, and that the choice of compression method, not merely its nominal bit-width, determines whether task accuracy survives deployment.
Training-free block-sparse attention can accelerate video transformers, but row-wise attention concentration does not by itself specify an executable sparse operator. Queries sharing a block route may have poorly overlapping supports, while retained attention mass alone does not determine the post-softmax error from skipped interactions. We show that partition geometry affects both pooled support and the predictability of the remaining residual from the sparse output. We introduce SparsePR, which combines Response-Coupled Partitioning with Probe-Fitted Residual Reconstruction. Sampled-query key responses form paired K/V groups, whose centroids induce query-response coordinates for shared routing. A small set of exact query rows then calibrates a call-specific affine correction from the sparse output within the output subspace observed in the probe residuals. Across four heterogeneous video generation and world models, SparsePR consistently reduces attention-reconstruction error. Ablations show that probe fitting accounts for most of this reduction, while response-coupled partitioning lowers hard-drop error and improves reconstruction under a finite probe budget. SparsePR preserves generation quality at 22.0-26.0% realized executed-pair density while achieving 1.48x-2.61x end-to-end speedups. Project page: https://pardistaghavi.github.io/SparsePR-website/
Diffusion models have achieved remarkable success in image and video generation, yet the high computational cost of iterative sampling remains a critical bottleneck for practical deployment. Feature caching has emerged as a promising acceleration paradigm by reusing or predicting intermediate features across timesteps. However, existing training-free methods apply uniform prediction strategies that cannot adapt to the heterogeneous feature dynamics, causing significant quality degradation under high acceleration ratios. We propose LinCa, a feature caching framework based on learnable invertible networks. LinCa decomposes cached features into sub-components with distinct continuity properties via a lightweight invertible network and applies differentiated prediction orders matched to each component. The strict invertibility guarantees lossless reconstruction back to the original feature space, forming a unified Decompose-Predict-Reconstruct pipeline. By training separate predictors for different models and timestep segments, LinCa adapts to heterogeneous feature dynamics. Experiments on FLUX, Qwen-Image, and HunyuanVideo demonstrate that LinCa, with less than 0.2% additional parameters, significantly outperforms existing methods and maintains near-lossless quality at 5-7x speedup. Code: https://github.com/QHR69/LinCa
Flow matching models for video generation achieve impressive performance but suffer from high computational overhead due to iterative denoising. In fact, the original model is not necessary for all denoising steps, allowing some steps to use lightweight alternatives for faster sampling. However, directly using caching or lightweight models can deviate from the original denoising trajectory, resulting in suboptimal performance. Through empirical analysis, we find that lightweight models can robustly capture the magnitude components of the original model's output, while caching provides reliable directional guidance. Building on this insight, we propose the Magnitude-Direction Decoupling (MDD) method, which adaptively employs a direction-calibrated lightweight model as a substitute for the original model to accelerate inference and effectively correct deviations in the denoising trajectory. Moreover, MDD further reduces inference costs by reusing magnitude information under classifier-free guidance (CFG). As a result, MDD offers a more reliable and lightweight solution to accelerate sampling. Experiments show that MDD outperforms existing acceleration methods, delivering promising speedups (e.g., up to 2.95x on Wan2.1) while preserving high visual fidelity and content richness.
Driving video generation models support autonomous-driving development by predicting controllable future scenes for simulation, planning evaluation, and offline data generation. Diffusion-based driving generators repeatedly evaluate large backbones across denoising steps, which limits generation throughput. Existing diffusion acceleration methods reduce this cost, but general-purpose designs omit driving signals available before generation, such as ego speed and planned trajectories. Experiments across driving motions show that cache tolerance varies with ego translation and rotation, denoising progress, and consecutive reuse length. We propose DriveCache, a training-free, action-aware controller that uses planned motion to allocate reuse across scenes and dynamic programming to place it across denoising steps under a calibrated response budget. A causal drift check refreshes features and replans the remaining schedule when generation departs from calibration. Across three generator configurations, DriveCache improves the overall fidelity-efficiency trade-off over evaluated cache methods. Our code will be publicly available.
Linear attention models eliminate the quadratic prefix computation and context-growing KV cache of softmax attention by replacing pairwise token interactions with recurrent state updates. However, existing decoding implementations often materialize and write back the full recurrent state after every generated token, making state maintenance a major source of memory traffic, especially for models with large states and many heads. This paper presents DeltaLog, a recurrent-state decoding scheme that reduces this overhead without changing the model semantics. Specifically, DeltaLog represents the recurrent state as a dense base state together with a bounded log of recent compact updates. Most decode steps append only compact update factors to this log, while periodic merge steps fold the accumulated updates back into the dense base state. Thus, the model observes the same dense state as in eager decoding, but most full-state write-backs are replaced by lightweight append operations. We implement DeltaLog for GDN, KDA, and RWKV6 and integrate it into a prototype serving stack. Across these models, DeltaLog accelerates the recurrent-state update kernel by up to $1.86\times$, reduces profiled recurrent-state write traffic by up to $7.83\times$, and achieves $1.05$--$1.20\times$ end-to-end serving speedups over dense recurrent baselines.