Efficient VLM Inference

VLM: Vision-Language Model

Ask about this topic

Overview

The field studies how to make vision-language models answer faster and use less memory and energy, especially for high-resolution images, long videos, and edge devices. Common approaches prune or merge visual tokens and frames, adapt resolution or retrieve regions on demand, reuse or compress KV caches, and use quantization, efficient kernels, routing, or speculative decoding. Methods are evaluated on tasks such as VQAv2, TextVQA, OCRBench, VideoMME, and LongVideoBench, measuring accuracy alongside time-to-first-token, throughput, and memory use. The central tension is preserving fine-grained evidence and reliable grounding while reducing computation: fewer tokens do not always mean lower end-to-end latency, and generating long answers can dominate cost on some devices.

Latest papers

CardsList

487 papers66 papers in the last four weeks, up 50% on the four weeks before. 0.4% of all new papers.

Apr 27, 2026

LiteVLA-H: Dual-Rate Vision-Language-Action Inference for Onboard Aerial Guidance and Semantic Perception

Vision-language-action (VLA) models have shown strong semantic grounding and task generalization in manipulation, but aerial deployment remains difficult because drones require low-latency closed-loop guidance under strict onboard compute and communication constraints. We present LiteVLA-H, a compact 256M-parameter VLA system designed for dual-rate operation on an NVIDIA Jetson AGX Orin: a fast outer-loop guidance mode for short action-token outputs and a slower semantic mode for scene understanding, hazard description, and operator-facing narration. The central empirical observation is that, in this compact edge regime, end-to-end latency is dominated by multimodal pre-fill rather than by the marginal cost of decoding a few extra tokens. This motivates a scheduler that issues reactive action tokens at 50.65,ms (19.74,Hz) while still supporting sentence-level semantic outputs at 149.90--164.57\ms (6.08--6.67,Hz) on the same embedded platform. To specialize the model without collapsing its descriptive competence, we use a knowledge-preserving fine-tuning recipe that mixes reactive flight data, aerial semantic data, and generic caption/VQA supervision. Beyond reporting current latency measurements, we position the system against recent state-of-the-art architectures, including AnywhereVLA, FutureVLA, and ReMem-VLA, showing that the measured action branch reaches a higher edge inference rate under our deployment conditions while retaining periodic semantic awareness.

Justin williams, Kishor Datta Gupta, Roy George +1

Aerial RoboticsEfficient VLM InferenceVision-Language-Action Models
AskPDF
Apr 27, 2026

SMoES: Soft Modality-Guided Expert Specialization in MoE-VLMs

Mixture-of-Experts (MoE) has become a prevalent backbone for large vision-language models (VLMs), yet how modality-specific signals should guide expert routing remains under-explored. Existing routing strategies are either hand-crafted or modality-agnostic, relying on idealized priors that ignore the layer-dependent modality fusion patterns in MoE-VLMs and provide little guidance for expert specialization. We propose Soft Modality-guided Expert Specialization (SMoES), which consists of dynamic soft modality scores that capture layer-dependent fusion patterns, an expert binning mechanism aligned with expert-parallel deployment, and an inter-bin mutual information regularization that encourages coherent modality specialization. Our method leverages attention-based or Gaussian-statistics modality scores to optimize mutual information regularization. Experiments across four MoE-based VLMs and 16 benchmarks demonstrate improvement on both effectiveness and efficiency: 0.9% and 4.2% average gain on multimodal and language-only tasks, 56.1% reduction in EP communication overhead, and 12.3% throughput improvement under realistic deployment. These results validate that aligning routing with modality-aware expert specialization unlocks MoE-VLM capacity and efficiency.

Zi-Hao Bo, Yaqian Li, Anzhou Hou +6

Li Auto Inc.

Vision-Language ModelsEfficient VLM InferenceMixture of ExpertsExpert RoutingMixture-of-Experts Models
AskPDF
Apr 27, 2026

LearnPruner: Rethinking Attention-based Token Pruning in Vision Language Models

Vision-Language Models (VLMs) have recently demonstrated remarkable capabilities in visual understanding and reasoning, but they also impose significant computational burdens due to long visual sequence inputs. Recent works address this issue by pruning unimportant visual tokens, achieving substantial computational reduction while maintaining model performance. The core of token pruning lies in determining token importance, with current approaches primarily relying on attention scores from vision encoders or Large Language Models (LLMs). In this paper, we analyze the effectiveness of attention mechanisms in both vision encoders and LLMs. We find that vision encoders suffer from attention sink, leading to poor focus on informative foreground regions, while in LLMs, although prior studies have identified attention bias toward token positions, text-to-vision attention demonstrates resistance to this bias and enables effective pruning guidance in middle layers. Based on these observations, we propose LearnPruner, a two-stage token pruning framework that first removes redundant vision tokens via a learnable pruning module after the vision encoder, then retains only task-relevant tokens in the LLM's middle layer. Experimental results show that our LearnPruner can preserve approximately 95% of the original performance while using only 5.5% of vision tokens, and achieve 3.2×\times inference acceleration, demonstrating a superior accuracy-efficiency trade-off.

Rinyoichi Takezoe, Yaqian Li, Zihao Bo +3

Li Auto Inc.

Efficient VLM InferenceAttention SinksVisual Token Pruning
AskPDF
Apr 27, 2026

GoClick: Lightweight Element Grounding Model for Autonomous GUI Interaction

Graphical User Interface (GUI) element grounding (precisely locating elements on screenshots based on natural language instructions) is fundamental for agents interacting with GUIs. Deploying this capability directly on resource-constrained devices like mobile phones is increasingly critical for GUI agents requiring low latency. However, this goal faces a significant challenge, as current visual grounding methods typically employ large vision-language model (VLM) (more than 2.5B parameters), making them impractical for on-device execution due to memory and computational constraints. To address this, this paper introduces GoClick, a lightweight GUI element grounding VLM with only 230M parameters that achieves excellent visual grounding accuracy, even on par with significantly larger models. Simply downsizing existing decoder-only VLMs is a straightforward way to design a lightweight model, but our experiments reveal that this approach yields suboptimal results. Instead, we select an encoder-decoder architecture, which outperforms decoder-only alternatives at small parameter scales for GUI grounding tasks. Additionally, the limited capacity of small VLMs encourages us to develop a Progressive Data Refinement pipeline that utilizes task type filtering and data ratio adjustment to extract a high-quality 3.8M-sample core set from a 10.8M raw dataset. Training GoClick using this core set brings notable grounding accuracy gains. Our experiments show that GoClick excels on multiple GUI element grounding benchmarks while maintaining a small size and high inference speed. GoClick also enhances GUI agent performance when integrated into a device-cloud collaboration framework, where GoClick helps cloud-based task planners perform precise element localization and achieve higher success rates. We hope our method serves as a meaningful exploration within the GUI agent community.

Hongxin Li, Yuntao Chen, Zhaoxiang Zhang

University of Chinese Academy of Sciences, Beijing, China. · New Laboratory of Pattern Recognition, State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing, China. · Hong Kong Institute of Science & Innovation, Chinese Academy of Sciences, Hong Kong, China.

Efficient VLM InferenceTraining Data CurationGUI AgentsGUI GroundingOn-Device Inference
AskPDF
Apr 26, 2026

Zoom In, Reason Out: Efficient Far-field Anomaly Detection in Expressway Surveillance Videos via Focused VLM Reasoning Guided by Bayesian Inference

Expressway video anomaly detection is essential for safety management. However, identifying anomalies across diverse scenes remains challenging, particularly for far-field targets exhibiting subtle abnormal vehicle motions. While Vision-Language Models (VLMs) demonstrate strong semantic reasoning capabilities, processing global frames causes attention dilution for these far-field objects and incurs prohibitive computational costs. To address these issues, we propose VIBES, an asynchronous collaborative framework utilizing VLMs guided by Bayesian inference. Specifically, to overcome poor generalization across varying expressway environments, we introduce an online Bayesian inference module. This module continuously evaluates vehicle trajectories to dynamically update the probabilistic boundaries of normal driving behaviors, serving as an asynchronous trigger to precisely localize anomalies in space and time. Instead of processing the continuous video stream, the VLM processes only the localized visual regions indicated by the trigger. This targeted visual input prevents attention dilution and enables accurate semantic reasoning. Extensive evaluations demonstrate that VIBES improves detection accuracy for far-field anomalies and reduces computational overhead, achieving high real-time efficiency and explainability while demonstrating generalization across diverse expressway conditions.

Xiaowei Mao, Bowen Sui, Weijie Zhang +7

School of Computer Science and Technology, Beijing Jiaotong University · 2Key Laboratory of Big Data & Artificial Intelligence in Transportation, Ministry of Education · 3Beijing Key Laboratory of Traffic Data Mining and Embodied Intelligence

Efficient VLM InferenceAnomaly LocalizationVideo Anomaly DetectionTraffic Anomaly DetectionAnomaly Detection
AskPDF
Apr 24, 2026

DocPrune:Efficient Document Question Answering via Background, Question, and Comprehension-aware Token Pruning

Recent advances in vision-language models have demonstrated remarkable performance across diverse multi-modal tasks, including document question answering that leverages structured visual cues from text, tables, and figures. However, unlike natural images, document images contain large backgrounds and only sparse supporting evidence, leading to the inefficient consumption of substantial computational resources, especially for long documents. We observe that existing token-reduction methods for natural images and videos fall short in utilizing the structural sparsity unique to documents. To address this, we propose DocPrune, a training-free and progressive document token pruning framework designed for efficient long-document understanding. The proposed method preserves only the essential tokens for the task while removing unnecessary ones, such as background or question-irrelevant tokens. Moreover, it automatically selects the appropriate layers to initiate token pruning based on the model's level of comprehension. Our experiments on the M3DocRAG show that DocPrune improves throughput by 3.0x and 3.3x in the encoder and decoder, respectively, while boosting the F1 score by +1.0, achieving both higher accuracy and efficiency without any additional training.

Joonmyung Choi, Sanghyeok Lee, Jongha Kim +4

Korea University · KAIST · Adobe Research

Efficient VLM InferenceDocument QAMultimodal Token CompressionVisual Token PruningDocument Image Understanding
AskPDF
Apr 23, 2026

Prototype-Based Test-Time Adaptation of Vision-Language Models

Test-time adaptation (TTA) has emerged as a promising paradigm for vision-language models (VLMs) to bridge the distribution gap between pre-training and test data. Recent works have focused on backpropagation-free TTA methods that rely on cache-based designs, but these introduce two key limitations. First, inference latency increases as the cache grows with the number of classes, leading to inefficiencies in large-scale settings. Second, suboptimal performance occurs when the cache contains insufficient or incorrect samples. In this paper, we present Prototype-Based Test-Time Adaptation (PTA), an efficient and effective TTA paradigm that uses a set of class-specific knowledge prototypes to accumulate knowledge from test samples. Particularly, knowledge prototypes are adaptively weighted based on the zero-shot class confidence of each test sample, incorporating the sample's visual features into the corresponding class-specific prototype. It is worth highlighting that the knowledge from past test samples is integrated and utilized solely in the prototypes, eliminating the overhead of cache population and retrieval that hinders the efficiency of existing TTA methods. This endows PTA with extremely high efficiency while achieving state-of-the-art performance on 15 image recognition benchmarks and 4 robust point cloud analysis benchmarks. For example, PTA improves CLIP's accuracy from 65.64% to 69.38% on 10 cross-domain benchmarks, while retaining 92% of CLIP's inference speed on large-scale ImageNet-1K. In contrast, the cache-based TDA achieves a lower accuracy of 67.97% and operates at only 50% of CLIP's inference speed.

Zhaohong Huang, Yuxin Zhang, Wenjing Liu +2

Key Laboratory of Multimedia Trusted Perception and Efficient Computing, Ministry of Education of China, Xiamen University, 361005, P.R. China.

Test-Time AdaptationVLM AdaptationTest-Time Adaptation of VLMsEfficient VLM Inference
AskPDF
Apr 23, 2026

Sub-Token Routing for KV Cache Compression

Transformer inference often requires a large KV cache, especially for long-context language modeling and multimodal generation. Existing compression methods usually reduce cache cost by selecting, evicting, quantizing, or compressing cached tokens, or by reducing the visual-token sequence before language-model inference. We introduce sub-token routing, a KV-compression method that adds a finer control axis inside retained tokens. It splits each retained value vector into groups and keeps only selected groups, while leaving query and key states unchanged. The method is designed to work after token-level reduction. First, a token-reduction method determines which tokens are retained. Then, sub-token routing compresses the value states inside those retained tokens. Experiments under matched KV budgets show that adding sub-token routing improves token-level reduction performance in both LLM and VLM settings, including Quest on LLaMA-2-7B and Qwen2.5-7B, and FastV/VisionZip across LLaVA and Qwen-VL models. The gains are larger at smaller KV budgets, suggesting that value-group routing is especially useful when further token removal becomes costly. Overall, token-level reduction and sub-token routing provide complementary ways to reduce KV cost.

Wei Jiang, Wei Wang

Futurewei Technologies San Jose, CA 95131, USA

LLM Inference EfficiencyKV CachingEfficient VLM InferenceKV-Cache CompressionToken Compression
AskPDF
Apr 22, 2026

Foveated Reasoning: Stateful, Action-based Visual Focusing for Vision-Language Models

Vision-language models benefit from high-resolution images, but the increase in visual-token count incurs high compute overhead. Humans resolve this tension via foveation: a coarse view guides "where to look", while selectively acquired high-acuity evidence refines "what to think". We introduce Foveated Reasoner, an autoregressive vision-language framework that unifies foveation and reasoning within a single decoding trajectory. Starting from a low-resolution view, the model triggers foveation only when needed, retrieves high-resolution evidence from selected regions, and injects it back into the same decoding trajectory. We train the method with a two-stage pipeline: coldstart supervision to bootstrap foveation behavior, followed by reinforcement learning to jointly improve evidence acquisition and task accuracy while discouraging trivial "see-everything" solutions. Experiments show that the method learns effective foveation policies and achieves stronger accuracy under tight visual-token budgets across multiple vision-language benchmarks.

Juhong Min, Lazar Valkov, Vitali Petsiuk +2

AI Center – Mountain View, Samsung Electronics

Visual ReasoningVisual AttentionEfficient VLM InferenceVLM ReasoningVisually Grounded ReasoningReasoning for VLMs+1
AskPDF
Apr 22, 2026

Sink-Token-Aware Pruning for Fine-Grained Video Understanding in Efficient Video LLMs

Video Large Language Models (Video LLMs) incur high inference latency due to a large number of visual tokens provided to LLMs. To address this, training-free visual token pruning has emerged as a solution to reduce computational costs; however, existing methods are primarily validated on Multiple-Choice Question Answering (MCQA) benchmarks, where coarse-grained cues often suffice. In this work, we reveal that these methods suffer a sharp performance collapse on fine-grained understanding tasks requiring precise visual grounding, such as hallucination evaluation. To explore this gap, we conduct a systematic analysis and identify sink tokens--semantically uninformative tokens that attract excessive attention--as a key obstacle to fine-grained video understanding. When these sink tokens survive pruning, they distort the model's visual evidence and hinder fine-grained understanding. Motivated by these insights, we propose Sink-Token-aware Pruning (SToP), a simple yet effective plug-and-play method that introduces a sink score to quantify each token's tendency to behave as a sink and applies this score to existing spatial and temporal pruning methods to suppress them, thereby enhancing video understanding. To validate the effectiveness of SToP, we apply it to state-of-the-art pruning methods (VisionZip, FastVid, and Holitom) and evaluate it across diverse benchmarks covering hallucination, open-ended generation, compositional reasoning, and MCQA. Our results demonstrate that SToP significantly boosts performance, even when pruning up to 90% of visual tokens.

Kibum Kim, Jiwan Kim, Kyle Min +4

Korea Advanced Institute of Science and Technology (KAIST) · Oracle · University of California, San Diego +1

Efficient VLM InferenceAttention SinksVisual Token Pruning
AskPDF
Apr 21, 2026

FOCAL: Filtered On-device Continuous Activity Logging for Efficient Personal Desktop Summarization

Desktop interaction streams provide a continuous, privacy-sensitive record of interleaved user tasks. Transforming these streams into task-organized personal logs on-device faces two main challenges: exhaustive Vision-Language Model (VLM) processing strains local resources, and global stream processing causes cross-task context pollution. We present FOCAL (Filtered On-device Continuous Activity Logging), a privacy-first multi-agent system utilizing a unified filter-plan-log architecture. It cascades a lightweight Filter Agent for noise suppression, a text-only Brain Agent for task attribution, a Record Agent for selective visual reasoning, and a task-isolated Memory Agent for context-coherent summarization. Experiments on DesktopBench (comprising 2,572 screenshots across 420 complex sessions) show FOCAL reduces total token consumption by 60.4% and VLM call count by 72.3% versus a baseline, while boosting Key Information Recall (KIR) from 0.38 to 0.61. Crucially, under A→B→AA{\to}B{\to}A task interruptions, FOCAL maintains Task Acc 0.81 and KIR 0.80, whereas the baseline collapses to Task Acc 0.03. FOCAL pioneers the efficient, on-device summarization of instruction-free desktop streams into multi-perspective personal logs.

Haoran Yin, Zhiyuan Wen, Jiannong Cao +2

The Hong Kong Polytechnic University Hong Kong, China · China Mobile Communications Company Limited Research Institute Beijing, China

Efficient VLM InferenceText SummarizationMemory-Augmented Language Model Agents
AskPDF
Apr 21, 2026

ST-Prune: Training-Free Spatio-Temporal Token Pruning for Vision-Language Models in Autonomous Driving

Vision-Language Models (VLMs) have become central to autonomous driving systems, yet their deployment is severely bottlenecked by the massive computational overhead of multi-view camera and multi-frame video input. Existing token pruning methods, primarily designed for single-image inputs, treat each frame or view in isolation and thus fail to exploit the inherent spatio-temporal redundancies in driving scenarios. To bridge this gap, we propose ST-Prune, a training-free, plug-and-play framework comprising two complementary modules: Motion-aware Temporal Pruning (MTP) and Ring-view Spatial Pruning (RSP). MTP addresses temporal redundancy by encoding motion volatility and temporal recency as soft constraints within the diversity selection objective, prioritizing dynamic trajectories and current-frame content over static historical background. RSP further resolves spatial redundancy by exploiting the ring-view camera geometry to penalize bilateral cross-view similarity, eliminating duplicate projections and residual background that temporal pruning alone cannot suppress. These two modules together constitute a complete spatio-temporal pruning process, preserving key scene information under strict compression. Validated across four benchmarks spanning perception, prediction, and planning, ST-Prune establishes new state-of-the-art for training-free token pruning. Notably, even at 90% token reduction, ST-Prune achieves near-lossless performance with certain metrics surpassing the full-model baseline, while maintaining inference speeds comparable to existing pruning approaches.

Lin Sha, Haiyun Guo, Tao Wang +4

School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100049, China · Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China · Carizon, Beijing 100094, China

VLMs for Autonomous DrivingEfficient VLM InferenceLarge Vision-Language ModelsAutonomous DrivingToken PruningVisual Token Pruning
AskPDF
Apr 20, 2026

Progressive Online Video Understanding with Evidence-Aligned Timing and Transparent Decisions

Visual agents operating in the wild must respond to queries precisely when sufficient evidence first appears in a video stream, a critical capability that is overlooked by conventional video LLMs evaluated in offline settings. The shift to an online, streaming paradigm introduces significant challenges: a lack of decision transparency, the difficulty of aligning response timing with visual evidence, and the need to maintain a global, causally consistent understanding under tight computational budgets. To address these issues, we propose a novel framework that decouples reasoning control from memory integration. We introduce \model{}, an instantiation of this framework with two core components. First, the Active Thinking Decision Maker (ATDM) is a transparent reasoning controller that externalizes its decision process using observable progress (ρ\boldsymbolρ) and confidence (c\boldsymbol{c}) metrics. This allows it to precisely time its response trt_r to match the first-sufficient-evidence timestamp t⋆t^\star while streaming its reasoning to the user. Second, the Hierarchical Progressive Semantic Integration (HPSI) module acts as an efficient memory system. It employs a set of learnable, multi-level aggregation tokens that are propagated across clips to build a rich, global cognitive state without exceeding token budgets. %Our approach sets a new standard on key online video understanding benchmarks, achieving strong performance of 71.6% on StreamingBench and 46.9% on OVOBench, demonstrating a robust solution for evidence-aligned and transparent online video analysis. Extensive experiments demonstrate the effectiveness of ATDM and HPSI, e.g., Thinking-QwenVL improves the accuracy of the previous state-of-the-art from 67.63% to 71.60% on the StreamingBench benchmark.

Kecheng Zhang, Zongxin Yang, Mingfei Han +6

University of Science and Technology of China · Harvard University · Department of CV, MBZUAI +3

Streaming Video UnderstandingEfficient VLM InferenceVideo QAVLM InterpretabilityMemory-Augmented Video UnderstandingMultimodal Understanding
AskPDF
Apr 20, 2026

Stability Implies Redundancy: Delta Attention Selective Halting for Efficient Long-Context Prefilling

Prefilling computational costs pose a significant bottleneck for Large Language Models (LLMs) and Large Multimodal Models (LMMs) in long-context settings. While token pruning reduces sequence length, prior methods rely on heuristics that break compatibility with hardware-efficient kernels like FlashAttention. In this work, we observe that tokens evolve toward semantic fixing points, making further processing redundant. To this end, we introduce Delta Attention Selective Halting (DASH), a training-free policy that monitors the layer-wise update dynamics of the self-attention mechanism to selectively halt stabilized tokens. Extensive evaluation confirms that DASH generalizes across language and vision benchmarks, delivering significant prefill speedups while preserving model accuracy and hardware efficiency. Code will be released at https://github.com/verach3n/DASH.git.

Yujie Chen, Tailai Chen, Yifeng Gao +4

Shanghai Jiao Tong University · University of California, San Diego · Carnegie Mellon University

Self-AttentionLong-Context Language Model InferenceEfficient VLM InferenceLLM Inference AccelerationAttention Mechanisms
AskPDF
Apr 19, 2026

Towards Joint Quantization and Token Pruning of Vision-Language Models

Deploying Vision-Language Models (VLMs) under aggressive low-bit inference remains challenging because inference cost is dominated by the long visual-token prefix during prefill and the growing KV cache during autoregressive decoding. Token pruning and low-bit quantization are complementary for reducing these costs, yet naive stage-wise combinations are often brittle due to a mismatch between quantization calibration and pruning execution. We present a collaborative quantization-and-pruning framework that unifies low-bit inference and deterministic visual-token pruning in a single deployable pipeline. The framework introduces the Quantization Unified Offline Token Allocator (QUOTA), which converts low-bit calibration signals into a layer-wise token allocation schedule and materializes it as a pruning recipe. Token importance is evaluated under deployed W4A4 operators with a quantized KV cache by combining activation magnitude, attention cues, and an explicit low-bit risk signal, enabling consistent budgeted top-kk selection. Experiments on standard VLM benchmarks show improved robustness over stage-wise baselines under the same low-bit regime, achieving 95.65% average retention while retaining only 30% of visual tokens, compared with about 94.3% retention for representative stage-wise combinations. The code will be released.

Xinqing Li, Xin He, Xindong Zhang +3

VCIP, College of Computer Science, Nankai University · School of Computer Science and Engineering, Tianjin University of Technology · OPPO Research Institute +3

Vision-Language ModelsVLM QuantizationEfficient VLM InferenceLow-Bit QuantizationVisual Token PruningNeural Network Compression+1
AskPDF
Apr 18, 2026

SinkRouter: Sink-Aware Routing for Efficient Long-Context Decoding in Large Language and Multimodal Models

In long-context decoding for LLMs and LMMs, attention becomes increasingly memory-bound because each decoding step must load a large amount of KV-cache data from GPU memory. Existing acceleration strategies often trade efficiency for accuracy by relying on heuristic pruning that may discard useful information. At a deeper level, they also tend to indiscriminately preserve all high-scoring tokens, treat early tokens as indispensable anchors, or rely on heuristic head routing, reflecting an insufficient mechanistic understanding of the attention sink phenomenon. In this paper, we show that the attention sink phenomenon corresponds to a stable, reachable, and error-controllable fixed point constructed during training. Based on this insight, we propose SinkRouter, a training-free selective routing framework that detects the sink signal and skips computations that would otherwise produce near-zero output. To translate this mechanism into real-world acceleration, we develop a hardware-aware Triton kernel with block-level branching and Split-K parallelism. We conduct extensive evaluations on a diverse suite of long-context benchmarks, including LongBench, InfiniteBench, CVBench, MileBench, and MMVP, using both text-only and multimodal backbones such as Llama-3.1-8B, Llama-3.1-70B, Yi-9B-200K, LLaVA-1.5-7B, and LLaVA-1.5-13B. Across these settings, SinkRouter consistently improves decoding efficiency while maintaining competitive accuracy, and reaches 2.03x speedup with a 512K context.

Junnan Liu, Xinyan Liu, Peifeng Gao +4

Harbin Institute of Technology (Weihai) · University of Hong Kong, Hong Kong, China · City University of Hong Kong, Hong Kong, China

GPU AccelerationSelf-AttentionLong-Context Language Model InferenceEfficient VLM InferenceLLM Inference AccelerationAttention Sinks+1
AskPDF
Apr 16, 2026

VisPCO: Visual Token Pruning Configuration Optimization via Budget-Aware Pareto-Frontier Learning for Vision-Language Models

Visual token pruning methods effectively mitigate the quadratic computational growth caused by processing high-resolution images and video frames in vision-language models (VLMs). However, existing approaches rely on predefined pruning configurations without determining whether they achieve computation-performance optimality. In this work, we introduce , a novel framework that formulates visual token pruning as a Pareto configuration optimization problem to automatically identify optimal configurations. Our approach employs continuous relaxation and straight-through estimators to enable gradient-based search, solved via the Augmented Lagrangian method. Extensive experiments across 8 visual benchmarks demonstrate that effectively approximates the empirical Pareto frontier obtained through grid search and generalizes well across various pruning methods and VLM architectures. Furthermore, through learnable kernel functions, we investigate layer-wise pruning patterns and reveal that multi-step progressive pruning captures VLMs' hierarchical compression structure, achieving superior accuracy-efficiency trade-offs compared to single-layer approaches.

Huawei Ji, Yuanhao Sun, Yuan Jin +4

Shanghai Jiao Tong University, Shanghai, China · University of Edinburgh, Edinburgh, UK

Vision-Language ModelsEfficient VLM InferenceMulti-Objective OptimizationVisual Token Pruning
AskPDF
Apr 16, 2026

Learning Adaptive Reasoning Paths for Efficient Visual Reasoning

Visual reasoning models (VRMs) have recently shown strong cross-modal reasoning capabilities by integrating visual perception with language reasoning. However, they often suffer from overthinking, producing unnecessarily long reasoning chains for any tasks. We attribute this issue to Reasoning Path Redundancy in visual reasoning: many visual questions do not require the full reasoning process. To address this, we propose AVR, an adaptive visual reasoning framework that decomposes visual reasoning into three cognitive functions: visual perception, logical reasoning, and answer application. It further enables models to dynamically choose among three response formats: Full Format, Perception-Only Format, and Direct Answer. AVR is trained with FS-GRPO, an adaptation of Group Relative Policy Optimization that encourages the model to select the most efficient reasoning format while preserving correctness. Experiments on multiple vision-language benchmarks show that AVR reduces token usage by 50--90% while maintaining overall accuracy, especially in perception-intensive tasks. These results demonstrate that adaptive visual reasoning can effectively mitigate overthinking in VRMs. Code and data are available at: https://github.com/RunRiotComeOn/AVR.

Yixu Huang, Tinghui Zhu, Muhao Chen

♠Fudan University · ♣University of California, Davis

Efficient VLM InferenceVisual ReasoningVLM ReasoningVisually Grounded ReasoningRL for VLMs
AskPDF
Apr 15, 2026

BARD: Bridging AutoRegressive and Diffusion Vision-Language Models Via Highly Efficient Progressive Block Merging and Stage-Wise Distillation

Autoregressive vision-language models (VLMs) deliver strong multimodal capability, but their token-by-token decoding imposes a fundamental inference bottleneck. Diffusion VLMs offer a more parallel decoding paradigm, yet directly converting a pretrained autoregressive VLM into a large-block diffusion VLM (dVLM) often leads to substantial quality degradation. In this work, we present BARD, a simple and effective bridging framework that converts a pretrained autoregressive VLM into a same-architecture, decoding-efficient dVLM. Our approach combines progressive supervised block merging, which gradually enlarges the decoding block size, with stage-wise intra-dVLM distillation from a fixed small-block diffusion anchor to recover performance lost at larger blocks. We further incorporate a mixed noise scheduler to improve robustness and token revision during denoising, and memory-friendly training to enable efficient training on long multimodal sequences. A key empirical finding is that direct autoregressive-to-diffusion distillation is poorly aligned and can even hurt performance, whereas distillation within the diffusion regime is consistently effective. Experimental results show that, with ≤\leq 4.4M data, BARD-VL transfers strong multimodal capability from Qwen3-VL to a large-block dVLM. Remarkably, BARD-VL establishes a new SOTA among comparable-scale open dVLMs on our evaluation suite at both 4B and 8B scales. At the same time, BARD-VL achieves up to 3×\times decoding throughput speedup compared to the source model. Code is available at https://github.com/fudan-generative-vision/Bard-VL.

Baoyou Chen, Hanchen Xia, Peng Tu +5

Shanghai Academy of AI for Science · Fudan University · Shanghai Innovation Institute +1

Vision-Language ModelsDiffusion VLMsVLM AdaptationDiffusion Model DistillationEfficient VLM InferenceVLM Distillation+1
AskPDF
Apr 13, 2026

Confidence under Visual Token Pruning: Removed Evidence and Risk-Controlled Token Budgets for MLLMs

Visual token pruning speeds up multimodal large language models (MLLMs) by keeping a small subset of the visual tokens, and pruning methods are compared by the accuracy they retain. We study what pruning does to the confidence of these models, across common selectors, several MLLMs, and different output formats. Pruning errors concentrate on questions whose evidence the selector removed, and confidence does not register the removal. When the queried object loses all of its tokens, accuracy on these questions drops from 59% to 17%, while confidence stays at the unpruned level. Returning a few object tokens to the kept set recovers most of the lost accuracy. Selectors that keep the most attended tokens remove such evidence most often and produce confident errors, which temperature scaling cannot re-rank. Selectors that avoid keeping redundant tokens stay close to the calibration of the unpruned model. We then use the confidence of the pruned model to set a per-question token budget. The model answers with few tokens first and again with all tokens when its confidence is low. Conformal risk control sets the threshold to bound the expected deviation from the unpruned model. With coverage-based selection, this cascade needs about a third of the prefill tokens of the unpruned model, while with FastV it needs more than the unpruned model. The savings come mainly from how well the confidence ranks the answers that differ from the unpruned ones.

Kaizhen Tan, Yang Feng, Heqing Du +3

Carnegie Mellon University, Pittsburgh, PA, USA · Columbia University, New York, NY, USA

VLM EvaluationConfidence Estimation for VLMsConfidence Estimation in Language ModelsEfficient VLM InferenceConformal Risk ControlVisual Token Pruning+1
AskPDF
Apr 13, 2026

POINTS-Long: Adaptive Dual-Mode Visual Reasoning in MLLMs

Multimodal Large Language Models (MLLMs) have recently demonstrated remarkable capabilities in cross-modal understanding and generation. However, the rapid growth of visual token sequences--especially in long-video and streaming scenarios--poses a major challenge to their scalability and real-world deployment. Thus, we introduce POINTS-Long, a native dual-mode MLLM featuring dynamic visual token scaling inspired by the human visual system. The model supports two complementary perception modes: focus mode and standby mode, enabling users to dynamically trade off efficiency and accuracy during inference. On fine-grained visual tasks, the focus mode retains the optimal performance, while on long-form general visual understanding, the standby mode retains 97.7-99.7% of the original accuracy using only 1/40-1/10th of the visual tokens. Moreover, POINTS-Long natively supports streaming visual understanding via a dynamically detachable KV-cache design, allowing efficient maintenance of ultra-long visual memory. Our work provides new insights into the design of future MLLMs and lays the foundation for adaptive and efficient long-form visual understanding.

Haicheng Wang, Yuan Liu, Yikun Liu +9

SAI, Shanghai Jiao Tong University, China · WeChat AI, Tencent, China

Multimodal Large Language ModelsLong-Context VLMsEfficient VLM InferenceVisual Token CompressionMemory-Augmented VLMs
AskPDF
Apr 13, 2026

Beyond Attention Scores: SVD-Based Vision Token Pruning for Efficient Vision-Language Models

Vision-Language Models (VLMs) have revolutionized multi-modal learning by jointly processing visual and textual information. Yet, they face significant challenges due to the high computational and memory demands of processing long sequences of vision tokens. Many existing methods rely on local heuristics, such as attention scores or token norms. However, these criteria suffer from positional bias and information dispersion, limiting their ability to preserve essential content at high pruning ratios and leading to performance degradation on visually detailed images. To address these issues, we propose SVD-Prune, a training-free, plug-and-play token pruning method based on Singular Value Decomposition. It decomposes the vision token feature matrix and selects the top-k tokens using statistical leverage scores, ensuring only tokens contributing most to the dominant global variance are preserved. Experiments show that SVD-Prune consistently outperforms prior pruning methods under extreme vision token budgets, maintaining strong performance even with 32 and 16 vision tokens.

Yvon Apedo, Martyna Poreba, Michal Szczepanski +1

Universit´e Paris-Saclay, Univ Evry, IBISC, 91020, Evry-Courcouronnes, France · Universit´e Paris-Saclay, CEA, List, F-91120, Palaiseau, France

Vision-Language ModelsEfficient VLM InferenceVisual Token Pruning
AskPDF
Apr 13, 2026

Decoupled Similarity for Task-Aware Token Pruning in Large Vision-Language Models

Token pruning has emerged as an effective approach to reduce the substantial computational overhead of Large Vision-Language Models (LVLMs) by discarding less informative visual tokens while preserving performance. However, existing methods typically rely on individual attention sources from different LVLM components, resulting in incomplete and suboptimal pruning decisions due to biased attention distributions. To address this problem, we propose DeSAP, a novel Decoupled Similarity-Aware Pruning method for precise, task-aware token pruning within the visual encoder. Specifically, DeSAP introduces a decoupled similarity to capture fine-grained cross-modal relevance between visual features and text tokens, providing explicit task-related guidance for pruning. By integrating decoupled similarity with visual saliency signals derived from visual attention, DeSAP performs token pruning under the guidance of both task-related and visual cues, enabling robust pruning even under aggressive pruning ratios. Extensive experiments across diverse benchmarks and architectures show that DeSAP consistently outperforms SOTA methods in both accuracy and efficiency. On LLaVA-1.5-7B, DeSAP achieves a 10 times FLOPs reduction and a 2.3 times prefill speedup by retaining only 11.1% of visual tokens, while maintaining 98.1% of the original performance.

Kexin Ma, Jing Xiao, Chaofeng Chen +4

School of Artificial Intelligence, Wuhan University, Wuhan, China · Department of Computer Science, University of Exeter, Exeter, United Kingdom · Chinese Academy of Sciences, Beijing, China; Wuhan AI Research, Wuhan, China +1

Visual Token PruningEfficient VLM Inference
AskPDF
Apr 13, 2026

Semantic-Geometric Dual Compression: Training-Free Visual Token Reduction for Ultra-High-Resolution Remote Sensing Understanding

Multimodal Large Language Models (MLLMs) have demonstrated immense potential in Earth observation. However, the massive visual tokens generated when processing Ultra-High-Resolution (UHR) imagery introduce prohibitive computational overhead, severely bottlenecking their inference efficiency. Existing visual token compression methods predominantly adopt static and uniform compression strategies, neglecting the inherent "Semantic-Geometric Duality" in remote sensing interpretation tasks. Specifically, object semantic tasks focus on the abstract semantics of objects and benefit from aggressive background pruning, whereas scene geometric tasks critically rely on the integrity of spatial topology. To address this challenge, we propose DualComp, a task-adaptive dual-stream token compression framework. Dynamically guided by a lightweight pre-trained router, DualComp decouples feature processing into two dedicated pathways. In the object semantic stream, the Spatially-Contiguous Semantic Aggregator (SCSA) utilizes size-adaptive clustering to aggregates redundant background while protecting small object. In the scene geometric stream, the Instruction-Guided Structure Recoverer (IGSR) introduces a greedy path-tracing topology completion mechanism to reconstruct spatial skeletons. Experiments on the UHR remote sensing benchmark XLRS-Bench demonstrate that DualComp accomplishes high-fidelity remote sensing interpretation at an exceptionally low computational cost, achieving simultaneous improvements in both efficiency and accuracy.

Yueying Li, Fengxiang Wang, Yan Li +3

College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China · School of Computer Science, Beijing University of Posts and Telecommunications, Beijing 100876, China

Visual Token CompressionEfficient VLM InferenceRemote Sensing Image Understanding
AskPDF
Apr 10, 2026

UIPress: Bringing Optical Token Compression to UI-to-Code Generation

UI-to-Code generation requires vision-language models (VLMs) to produce thousands of tokens of structured HTML/CSS from a single screenshot, making visual token efficiency critical. Existing compression methods either select tokens at inference time using task-agnostic heuristics, or zero out low-attention features without actually shortening the sequence -- neither truly reduces prefill latency or adapts to the non-uniform information density of UI screenshots. Meanwhile, optical (encoder-side learned) compression has shown strong results for document OCR, yet no prior work has adapted this paradigm to UI-to-Code generation. We propose UIPress, a lightweight learned compression module inserted between the frozen ViT encoder and the LLM decoder of Qwen3-VL-8B. UIPress combines depthwise-separable convolutions, element-guided spatial reweighting, and Transformer refinement to compress ∼{\sim}6{,}700 visual tokens to a fixed budget of 256. Together with Low-Rank Adaptation (LoRA) on the decoder to bridge the representation gap, the entire system adds only ∼{\sim}21.7M trainable parameters (0.26% of the 8B base model). Under a fair comparison on the same base model against four baselines on Design2Code, UIPress at 256 tokens achieves a CLIP score of 0.8127, outperforming the uncompressed baseline by +7.5% and the strongest inference-time method by +4.6%, while delivering 9.1×\times time-to-first-token speedup. To the best of our knowledge, UIPress is the first encoder-side learned compression method for the UI-to-Code task.

Dasen Dai, Shuoqi Li, Ronghao Chen +3

The Chinese University of Hong Kong China · Peking University China · University of Chinese Academy of Sciences China +2

Token CompressionVisual Token CompressionImage-to-Code GenerationEfficient VLM InferenceCode GenerationVision-Language Models
AskPDF
Apr 10, 2026

How Should Video LLMs Output Time? An Analysis of Efficient Temporal Grounding Paradigms

While Multimodal Large Language Models (MLLMs) have advanced Video Temporal Grounding (VTG), existing methods often couple output paradigms with different backbones, datasets, and training protocols. This makes it challenging to isolate the specific impact of the output design. Additionally, as VTG systems are increasingly considered for resource-constrained edge deployment, the trade-off between output formulation and system-level efficiency requires systematic investigation. In this paper, we present a controlled empirical study comparing three dominant VTG output paradigms: Text Numeral Generation, Temporal Token Generation, and Continuous Temporal Decoding. We evaluate these paradigms across identical compact VLMs (SmolVLM2, FastVLM, and Molmo2) using consistent datasets and LoRA fine-tuning protocols. Evaluations on Charades-STA, QVHighlights, and YouCook2 measure both localization accuracy and system efficiency, including inference latency, training throughput, and parameter overhead. Our results demonstrate that the choice of output formulation significantly affects both grounding accuracy and computational cost, independent of model scale. Specifically, the continuous distribution paradigm consistently achieves the most favorable efficiency-accuracy trade-off on the Pareto frontier, delivering robust localization with minimal latency overhead. These findings provide objective empirical guidelines for designing efficient, deployment-ready VTG systems.

Shengji Jin, Yuanhao Zou, Victor Zhu +2

The University of Central Florida · Axon

Temporal Video GroundingEfficient VLM InferenceVideo-Language Model Evaluation
AskPDF
Apr 9, 2026

SMART: When is it Actually Worth Expanding a Speculative Tree?

Tree-based speculative decoding accelerates autoregressive generation by verifying a branching tree of draft tokens in a single target-model forward pass. However, existing methods prioritize maximizing token-level likelihood or the number of accepted tokens while ignoring a critical ``efficiency paradox'': the computational overhead of drafting and verifying big trees can grow super-linearly, particularly at scale. This often leads to negative wall-clock speedup when batch sizes increase or hardware saturation limits are reached. To address this, we propose SMART, a system-aware marginal analysis framework for runtime tree construction. SMART reformulates tree expansion as a hardware-aware optimization problem that directly maximizes end-to-end speedup. By applying a principled marginal benefit--cost rule at inference time, SMART expands a node only when its marginal benefit--cost ratio exceeds the tree-level speedup. SMART is training-free and serves as a plug-and-play controller for existing frameworks like MSD and EAGLE. Extensive evaluations across three MLLMs (e.g., LLaVA, Qwen2-VL) and four LLMs (e.g., Llama-3.1, DeepSeek-R1) demonstrate that SMART consistently outperforms state-of-the-art baselines. It delivers an average additional speedup of 20.0% for MLLMs and 15.4% for LLMs across compute-bound batching regimes and diverse GPU architectures without performance loss.

Lifu Wang, Pan Zhou

Singapore Management University, Singapore

Inference-Time OptimizationGPU AccelerationEfficient VLM InferenceSpeculative DecodingLLM Inference Acceleration
AskPDF
Apr 9, 2026

AdaSpark: Adaptive Sparsity for Efficient Long-Video Understanding

Processing long-form videos with Video Large Language Models (Video-LLMs) is computationally prohibitive. Current efficiency methods often compromise fine-grained perception through irreversible information disposal or inhibit long-range temporal modeling via rigid, predefined sparse patterns. This paper introduces AdaSpark, an adaptive sparsity framework designed to address these limitations. AdaSpark first partitions video inputs into 3D spatio-temporal cubes. It then employs two co-designed, context-aware components: (1) Adaptive Cube-Selective Attention (AdaS-Attn), which adaptively selects a subset of relevant video cubes to attend for each query token, and (2) Adaptive Token-Selective FFN (AdaS-FFN), which selectively processes only the most salient tokens within each cube. An entropy-based (Top-p) selection mechanism adaptively allocates computational resources based on input complexity. Experiments demonstrate that AdaSpark significantly reduces computational load by up to 57% FLOPs while maintaining comparable performance to dense models and preserving fine-grained, long-range dependencies, as validated on challenging hour-scale video benchmarks.

Handong Li, Zikang Liu, Longteng Guo +10

School of Artificial Intelligence, University of Chinese Academy of Sciences · Institute of Automation, Chinese Academy of Sciences · Alibaba Group Holding Limited +1

Long-Video UnderstandingEfficient VLM InferenceSparse AttentionConditional Computation
AskPDF
Apr 9, 2026

HAWK: Head Importance-Aware Visual Token Pruning in Multimodal Models

In multimodal large language models (MLLMs), the surge of visual tokens significantly increases the inference time and computational overhead, making them impractical for real-time or resource-constrained applications. Visual token pruning is a promising strategy for reducing the cost of MLLM inference by removing redundant visual tokens. Existing research usually assumes that all attention heads contribute equally to the visual interpretation. However, our study reveals that different heads may capture distinct visual semantics and inherently play distinct roles in visual processing. In light of this observation, we propose HAWK, a head importance-aware visual token pruning method that perceives the varying importance of attention heads in visual tasks to maximize the retention of crucial tokens. By leveraging head importance weights and text-guided attention to assess visual token significance, HAWK effectively retains task-relevant visual tokens while removing redundant ones. The proposed HAWK is entirely training-free and can be seamlessly applied to various MLLMs. Extensive experiments on multiple mainstream vision-language benchmarks demonstrate that HAWK achieves state-of-the-art accuracy. When applied to Qwen2.5-VL, HAWK retains 96.0% of the original accuracy after pruning 80.2% of the visual tokens. Additionally, it reduces end-to-end latency to 74.4% of the original and further decreases GPU memory usage across the tested models. The code is available at https://github.com/peppery77/HAWK.git.

Qihui Zhu, Tao Zhang, Yuchen Wang +9

University of Science and Technology of China · ChangXin Memory Technologies, Inc · Huawei Noah’s Ark Lab

Visual Token PruningEfficient VLM InferenceLarge Vision-Language Models
AskPDF
Apr 8, 2026

Q-Zoom: Query-Aware Adaptive Perception for Efficient Multimodal Large Language Models

MLLMs require high-resolution visual inputs for fine-grained tasks like document understanding and dense scene perception. However, current global resolution scaling paradigms indiscriminately flood the quadratic self-attention mechanism with visually redundant tokens, severely bottlenecking inference throughput while ignoring spatial sparsity and query intent. To overcome this, we propose Q-Zoom, a query-aware adaptive high-resolution perception framework that operates in an efficient coarse-to-fine manner. First, a lightweight Dynamic Gating Network safely bypasses high-resolution processing when coarse global features suffice. Second, for queries demanding fine-grained perception, a Self-Distilled Region Proposal Network (SD-RPN) precisely localizes the task-relevant Region-of-Interest (RoI) directly from intermediate feature spaces. To optimize these modules efficiently, the gating network uses a consistency-aware generation strategy to derive deterministic routing labels, while the SD-RPN employs a fully self-supervised distillation paradigm. A continuous spatio-temporal alignment scheme and targeted fine-tuning then seamlessly fuse the dense local RoI with the coarse global layout. Extensive experiments demonstrate that Q-Zoom establishes a dominant Pareto frontier. Using Qwen2.5-VL-7B as a primary testbed, Q-Zoom accelerates inference by 2.52 times on Document & OCR benchmarks and 4.39 times in High-Resolution scenarios while matching the baseline's peak accuracy. Furthermore, when configured for maximum perceptual fidelity, Q-Zoom surpasses the baseline's peak performance by 1.1% and 8.1% on these respective benchmarks. These robust improvements transfer seamlessly to Qwen3-VL, LLaVA, and emerging RL-based thinking-with-image models. Project page is available at https://yuhengsss.github.io/Q-Zoom/.

Yuheng Shi, Xiaohuan Pei, Linfeng Wen +2

University of Sydney, Australia · Sun Yat-sen University, China · City University of Hong Kong, China

Efficient VLM InferenceLarge Vision-Language Models
AskPDF