Overview

Visual token pruning studies how to remove or merge image and video patches before or inside vision transformers and vision-language models, reducing computation and memory for tasks such as visual question answering, video understanding, document reading, and robot control. Methods use attention or text-query relevance, feature diversity and spatial coverage, and signals such as entropy, spectral structure, or geometry; learned selectors and adaptive budgets choose different tokens or compression rates for different inputs. Pruning can happen before visual encoding, across language-model layers, or during video processing, and is evaluated on both task benchmarks and practical measures such as prefill latency, FLOPs, and memory. The central tension is that a token that seems redundant early may become important for later reasoning, while selection overhead and hardware execution can prevent fewer tokens from yielding faster end-to-end inference.

Latest papers

295 papers23 papers in the last four weeks, down 48% on the four weeks before. 0.2% of all new papers.

Oct 14, 2026

STP-GUI: Spatially-Aware Token Pruning for Efficient Web-GUI Agents

With the rapid development of vision-language models (VLMs), Web-GUI agents have achieved promising performance in interacting with real-world interfaces through screenshot-based grounding. However, high-resolution screenshots and multi-turn histories introduce excessive visual tokens, causing substantial inference overhead. Existing token pruning methods, mainly designed for natural images, often select tokens independently based on importance scores, which can fragment the local visual structures required for recognizing small GUI elements such as buttons, icons, and hyperlinks. In this work, we propose STP-GUI (Spatially-aware Token Pruning for GUI agents), a query-guided token pruning framework that preserves fine-grained instruction-conditioned token scoring while performing spatially coherent window-level selection under aggressive token reduction. A lightweight selector is optimized through policy learning while the VLM backbone remains frozen. Experiments on Online-Mind2Web show that \papername achieves a favorable efficiency-performance tradeoff under low token budgets. At 21% token retention, it preserves 77.7% of full-token performance while reducing inference latency by 31% and FLOPs by 57%. Results on Android in the Wild further suggest that the method transfers to mobile GUI environments.
Oct 14, 2026

Complementary Future Token Pruning for World Action Models

World action models (WAMs) jointly predict future visual states and robot actions, but processing dense future-frame tokens incurs substantial inference latency. We propose CF-Prune, a training-free framework for pruning future-frame tokens in pretrained WAMs. Our key insight is that future visual predictions support action generation and therefore need not preserve every visual detail required for high-fidelity video synthesis. CF-Prune combines instruction and action attention to identify task-relevant tokens during the initial denoising forward pass, even when future-frame and action representations remain highly noisy. An entropy-based phase identification mechanism further retains dense computation during precision phases to mitigate pruning-induced performance degradation. Experiments on RoboLab, RoboCasa, and real-world manipulation tasks demonstrate consistent inference speedups. At a 35% pruning ratio, CF-Prune achieves average speedups of 1.19-1.27x while improving task success rates. At a 50% pruning ratio, it achieves 1.29-1.33x average speedups on both simulation benchmarks, with overall success rates within 1.0 percentage point of the baseline.
Oct 13, 2026

Just Respect the Semantics: Image-Token Compression via Pre-LLM Semantic Variable-Resolution Tokenization and In-LLM Relevance Pruning

Image-token compression must reduce computation without hurting the semantics. We introduce Varest, guided by semantic fungibility: tokens within a coherent semantic region can be summarized coarsely, while object boundaries and literal text require finer distinctions. Object and text localizers guide variable-resolution merging, preserving spatial coverage and finer detail at boundaries and text. Learned [REL] registers then prune visual tokens after early decoder processing. Training across budgets enables one checkpoint per backbone. Across three backbones and 13 benchmarks, Varest outperforms 15 baselines in mean performance retention at every evaluated budget, retaining 86.10-92.33% of full-token performance with only 5.56% of the original layer-averaged image-token count.
Oct 13, 2026

Bifocal: Training-Free Mixed Granularity Visual Token Pruning for Multimodal LLMs

Visual token pruning for multimodal large language models (MLLMs) appears highly effective, with existing methods retaining over 95% of unpruned performance on common VQA benchmarks using a quarter of the visual tokens or fewer. However, our extensive evaluation of ten methods across three MLLM families and five ability groups reveals a critical gap: strong general performance often comes at the cost of severe degradation on fine-grained tasks such as OCR and visual grounding. We trace these failures to two limitations of the prevailing paradigm of retaining the highest-ranked native patches. A single importance ranking overlooks the complementary evidence required by different tasks, while sparse patch selection disrupts both image content and spatial integrity. To address these limitations, we propose Bifocal, a training-free, two-stage pruning method built on a mixed-granularity framework. It allocates separate budgets to query-focused and image-intrinsic evidence, complemented by a pooled coarse representation of the entire image. Before the LLM, Bifocal matches question words to visual regions through a soft conjunction balanced against norm-debiased image saliency, while selecting image-intrinsic evidence to maximize representational coverage. Inside the LLM, it further prunes tokens using density-adjusted text-to-image attention. With roughly a quarter of the original visual tokens, Bifocal is the only evaluated method to retain over 90% of unpruned performance in every ability group across all three models, and ranks first on average in all nine model-budget settings. The code is available at: https://github.com/EIT-NLP/Bifocal
Oct 13, 2026

Prune-and-Merge: Training-Free Scaling of Vision Transformers

Vision Transformers achieve strong performance but incur high computational costs from quadratic token interactions. Training-free token reduction methods improve efficiency, but their robustness remains underexplored. We propose Prune-and Merge ViT, a training-free framework for off-the-shelf models. It combines class attention-based pruning with similarity-based merging to reduce redundancy while preserving semantic information. To evaluate robustness we introduce two metrics. Relative Accuracy Drop (RAD) measures relative accuracy degradation under transformations, and Prediction Flip Rate (PFR) quantifies prediction consistency. On pretrained ViT and DeiT, Prune-and-Merge reduces computation competitively while maintaining robustness and better accuracy than existing token reduction methods. Our analysis suggests that hybrid token reduction does not introduce hallucination-like behavior and preserves prediction stability for common image transformations.
Oct 13, 2026

How Much Can We Prune? Exploring the Performance Limits of Visual Token Pruning

Visual token pruning is increasingly used to reduce the memory and inference costs of vision-language models. Existing work has largely focused on designing method-specific selectors and demonstrating incremental improvements over prior methods at selected token budgets. It remains unclear whether the resulting performance reflects an inherent limit of visual token pruning or the limitations of current token-selection algorithms. We take a complementary perspective and ask: how well can visual token pruning perform when offline search cost is not the primary constraint? We formulate pruning for general image understanding as selecting, under a fixed budget, one shared visual-token subset that supports a broad distribution of questions about an image. We approximate this utility with a finite auxiliary QA set generated by a frontier VLM and optimize a teacher-forced answer-probability objective with the evaluated model frozen. Motivated by the conceptual monotonicity and diminishing returns of adding visual evidence, we construct nested subsets greedily and introduce QA mini-batch greedy (QA-MBG), which estimates token marginal gains from randomly sampled QA components at each step. Empirically, QA-MBG is both cheaper and more effective than full greedy. With one frontier-VLM call and approximately 0.50.5 H100 GPU-hours per image, it produces a nested subset trajectory covering keep ratios up to 0.30.3. Across Qwen3-VL-4B and Gemma~4 E4B on 900900 images from nine benchmarks, we find significant headroom over evaluated pruning baselines on held-out QA and image-grounded long-form generation, while the searched subsets substantially narrow the gap to full-token inference.
Oct 13, 2026

TwinPrune: Density-Aware Two-Phase Token Pruning for Vision-Language Models

We propose a unified spatial-density signal to score token importance for both inference phases: bipartite merge for prefill and KV eviction for decode, yielding measurable gains in each stage. To isolate the two phases, we split benchmarks by scoring logic. Prefill probes (MMBench, POPE, ScienceQA) evaluate only the first decoded token; their accuracy reflects prefill logits argmax before any decode KV eviction. Full-pipeline probes (TextVQA, DocVQA, MM-Vet) assess the complete generated sequence and capture both phases. We identify a critical evaluation pitfall: decode pruning claims validated solely on prefill probes measure prefill quality, not decode performance. Experiments show discarding all visual KV during decode barely affects prefill-probe accuracy but degrades TextVQA and MM-Vet drastically; this insensitivity stems from the scoring rule, not compression intensity. Evaluated on seven model variants (InternVL-3.5, Qwen3-VL, LLaVA-1.5) and three full-pipeline benchmarks, our method achieves a three-way Pareto frontier across compute, memory and accuracy. Density-based prefill compression maintains strong accuracy, scales linearly with visual token count, and delivers 100x faster token selection at high resolution. For decode, density top-k KV eviction consistently outperforms strong baselines H2O, SnapKV and StreamingLLM on TextVQA and MM-Vet across three retention ratios. At 0.125 keep ratio on TextVQA, it surpasses H2O by 9.6 points; InternVL-3.5 models (2B to 38B) maintain a 7 to 10 point advantage. Compression scales favorably with model size: every measured 8B to 38B model transition reduces accuracy loss instead of worsening it. We suggest MM-Vet or DocVQA as mandatory full-pipeline validation for any decode-phase claims. Our code is available at https://github.com/Kai-Liu001/TwinPrune
Oct 12, 2026

FoveaTree: Information-Density Trees for Query-Guided Visual Token Pruning in Video Large Language Models

Video large language models (Video LLMs) encode each frame into hundreds of visual tokens, making video inference expensive. Existing methods reduce this cost through similarity-based merging or attention-based selection but face two challenges: Hierarchy-blind Pruning, which overlooks regional compressibility and can discard critical evidence, and Attention Drift, where positional effects and attention sinks shift text-to-vision attention away from task relevance. We propose FoveaTree, a two-stage, training-free visual-token pruning framework. First, Merge-Gain Density Estimation measures the feature distortion caused by regional merging and constructs a quadtree-derived density map to guide token dropping, merging, and retention. Second, Attention Tree Pruning uses the surviving units and inherited tree topology to refine text-to-vision attention in intermediate LLM layers, preserving relevant regional evidence while removing low-relevance content. We evaluate FoveaTree on InternVL3.5-8B and Qwen3-VL-8B across five video question-answering benchmarks and one captioning benchmark at 50%, 30%, and 10% visual-token retention. FoveaTree consistently achieves overall state-of-the-art performance across the evaluated settings, offering a strong accuracy-efficiency trade-off.
Oct 12, 2026

When 1+1 Backfires: Rethinking Relevance-Diversity Balancing for Sample-Aware Visual Token Reduction

Visual token compression has emerged as a critical technique for reducing the inference cost of Vision-Language Models. Existing methods select tokens based on text relevance, visual diversity, or a fixed combination of both---treating the relevance-diversity balance as a universal constant independent of the input. We conduct a systematic analysis to challenge this assumption. At the dataset level, grid-searching the optimal relevance-diversity weight per benchmark reveals substantial cross-task variation. More critically, we further maximize per-sample answer likelihood to recover sample-level oracle weights, finding that their distributions not only differ markedly across benchmarks but also exhibit high within-benchmark variance. Based on these observations, we propose SAVER, a framework for Sample-Aware relevance-diversity balanced Visual tokEn Reduction. We first derive a Geometric Intent Prior (GIP) from the spatial concentration of text-relevant tokens to infer a sample-specific relevance-diversity weight, and then apply this weight to jointly select tokens via a Weighted Text Centroid (WTC) for relevance scoring and a Visual Marginal Coverage Gain (VMCG) for diversity. SAVER requires no additional training and achieves sample-aware balancing entirely at inference time. Extensive experiments demonstrate that SAVER consistently outperforms state-of-the-art baselines across diverse benchmarks, retaining 96.5% of full-model performance on LLaVA-1.5-7B with only 11.1% of the original visual tokens, and further generalizing to LLaVA-NeXT and Mini-Gemini.
Oct 9, 2026

SheafPrune: Sheaf-Guided Graph Propagation for Visual Token Pruning

Multimodal large language models process the complete visual token sequence, increasing prefill cost even when many tokens are irrelevant to the instruction. Token-wise saliency captures limited inter-token dependence; graph-based selection instead depends on how relations govern importance propagation and ranking. We study support-constrained propagation with node-degree correction. Inspired by sheaf geometry, SheafPrune uses modality-specific maps and a sparse text--visual graph to transfer relevance, then propagates scores on a separate visual graph and corrects them for degree. It retains a hard top-kk subset once before the frozen language model. The selector has about 4.54.5M parameters and is trained end-to-end through the frozen LLaVA-NeXT-8B backbone using only the language-modeling objective. Explicit retention-ratio conditioning lets one checkpoint serve every budget. Across four backbones, ten benchmarks and seven pruning ratios, with all baselines re-implemented under one protocol, SheafPrune achieves the best retention at 2222 of 2828 backbone--budget settings, with monotonically growing margins in the moderate-to-aggressive regime. On LLaVA-NeXT, retaining 10%10\% of visual tokens yields a 5.15×5.15\times net prefill speedup and an 84.0%84.0\% end-to-end FLOPs reduction. Together with component ablations, these results support graph-constrained importance propagation and degree correction as a design for visual token selection across the evaluated backbones and tokenization schemes, improving the trade-off between accuracy retention and inference cost. code:https://github.com/Bob0817912/SheafPrune_ICLR_2027
Oct 8, 2026

Look Back, Think Ahead: Visual Memory on Demand for Efficient Multimodal Reasoning

Processing long visual token sequences from high-resolution images makes multi-step reasoning computationally expensive for multimodal Large Language Models (MLLMs). Existing one-shot pruning and aggregation methods compress visual tokens into a fixed context before decoding. However, visual evidence needs can shift as reasoning unfolds, making it difficult for a fixed compressed context to retain all the details needed across stages. To address this challenge, we propose ViMoD, a lightweight framework that maintains a compact visual context while preserving access to original fine-grained evidence as reasoning needs evolve. Deformable Aggregation of Region-wise Tokens (DART) learns content-adaptive groups and aggregation capacities, constructing compact Coarse representations linked to recoverable original Fine tokens. Temporal Routing for Adaptive Contextual Evidence (TRACE) integrates decoding history to anticipate upcoming evidence needs and select, retain, or replace active Fine-token groups. Selected Fine tokens augment the persistent Coarse context in the frozen backbone, enabling stage-specific evidence access without continuously attending to all visual tokens. On Qwen3-VL-4B, ViMoD outperforms all evaluated baselines on all eight reasoning benchmarks at a 20% target visual token budget, improving the mean normalized score by 39.0% over the strongest evaluated one-shot baseline. These gains are achieved with only 0.0546% additional trainable parameters relative to the frozen backbone.
Oct 7, 2026

Understanding and Mitigating Token-Pruning-Induced Vulnerabilities in VLMs

Token-Pruning accelerates Vision-Language Models by removing redundant visual tokens, yet its safety implications remain underexplored. In this work, we present the first comprehensive safety evaluation of Token-Pruning mechanisms and find that: most pruning strategies significantly degrade safety as pruning ratios increase, whereas Query-based Compression shows the opposite, with extreme pruning (up to 99.8%), unexpectedly improves model safety. This sharp contrast prompts a key question: How do different Token-Pruning strategies reshape model safety behavior, and is it possible to enhance safety without sacrificing acceleration? To answer this, we identify an unrecognized mechanism, termed Pruning-Induced Malicious Amplification, where removal of background tokens triggers a side effect: forcing the model's attention to collapse onto a few retained malicious anchors within the foreground, inadvertently amplifying their toxic semantics under jailbreak. To address that, we propose an inference-time and plug-and-play Safety-Aware Pruning (SAP) mechanism that counteracts such dominance via three steps: (1) identifying malicious anchors, (2) restoring pruned benign tokens, and (3) reallocating excessive attention from malicious anchors to benign tokens. Extensive experiments across three safety and four utility benchmarks demonstrate that SAP mitigates pruning-induced vulnerabilities, i.e., reducing ASR by up to 62%, without compromising efficiency or utility.
Oct 7, 2026

Adaptive Visual Token Reduction for Accelerated Image Understanding

Large Vision-Language Models achieve strong VQA performance, but processing high-resolution, information-rich images requires substantial computation, motivating visual token reduction. However, existing methods often prune individual tokens or rely on fixed-size cropping, limiting their ability to preserve spatially structured information such as horizontally or vertically elongated text. To address this limitation, we propose ReFIT, an instruction-guided visual token reduction framework for efficient LVLM inference. ReFIT consists of Relevance-Guided Window Reshaping (RWR) and Instruction-Guided Token Refinement (ITR), where RWR captures instruction-relevant regions by adapting to their spatial characteristics, while ITR further removes unnecessary visual tokens. Experiments on four VQA benchmarks demonstrate that ReFIT improves answer accuracy while reducing computational cost, and qualitative results demonstrate its effectiveness in localizing relevant regions and removing unnecessary visual information.
Oct 6, 2026

DIPrune: Task-Aware Token Pruning with Dual Importance for Efficient Multimodal Language Models

Recent training-free pruning approaches for Multimodal Large Language Models (MLLMs) effectively cut computational overhead by exploiting visual redundancy or text-vision attention. However, they frequently suffer from semantic degradation due to their task-agnostic design or unreliable attention estimates. Based on our empirical analysis, we have found that this issue arises because salient tokens in shallow layers persistently suppress emerging semantic ones through numerical inertia, leading to premature discarding of signals crucial for deep reasoning. To address the aforementioned issue, from the task-oriented aspects, we first reformulate training-free pruning as a minimization of the distortion in the final task loss and derive a tractable, token-wise upper bound to serve as a surrogate objective. Specifically, this formulation inherently reveals a previously neglected inter-layer term that accounts for gradients across layers. Accordingly, for the implementation, we propose DIPrune, a rank-based framework that employs a dual importance scoring mechanism to jointly optimize intra-layer static feature saliency and inter-layer dynamic semantic evolution. Extensive experiments on LLaVA and Qwen-VL demonstrate that DIPrune consistently achieves state-of-the-art results.
Oct 6, 2026

VLA-ACL: Action-Consistent Visual Token Pruning for Efficient Vision-Language-Action Models

Vision-Language-Action (VLA) models achieve strong robotic manipulation performance but incur high computational costs from processing long token sequences at every control step, limiting real-time deployment. Visual token pruning offers a direct solution, as visual patches dominate the input sequence and contain considerable redundancy. Existing approaches, however, either rely on indirect training-free heuristics, such as attention scores and motion thresholds, or require costly fine-tuning of the base VLA model. We introduce VLA-ACL (Action Consistency Learning), which learns a lightweight visual token pruning policy through action-level supervision while keeping the base VLA model entirely frozen. The training objective encourages actions produced from pruned visual contexts to remain consistent with the full-context teacher, with ground-truth actions as auxiliary supervision. This directly ties token selection to its effect on the downstream control output. Experiments on LIBERO and real-world manipulation tasks show that VLA-ACL prunes up to 87.5% of visual tokens while retaining competitive performance, reduces computation by up to 75%, and achieves a 1.5x inference speedup. These results establish a stronger performance-efficiency trade-off than existing frozen-VLA pruning methods and demonstrate the value of action-level supervision for visual token selection. Code is available at https://github.com/du-owen/VLA-ACL.
Oct 6, 2026

Later Is Better: Token Reduction for ViTs Under Distribution Shift

Training-free token reduction accelerates vision transformers by removing redundant tokens across layers, recovering most of the original accuracy at a fraction of the compute. These methods, however, are designed and evaluated primarily on clean data, and under real-world distribution shift their accuracy gap to the uncompressed model widens with the removal rate. We show that this gap is governed by the reduction schedule, the depth profile of removal, usually left fixed as an implementation detail. Concretely, we introduce a one-parameter late-concentrated power-law schedule that consistently improves out-of-distribution accuracy over flat at no extra inference cost. On ImageNet-C with DeiT-S, the late schedule closes 83% of that gap at a 26% compute reduction (+1.17pp), and 99% of it at a lighter 7% reduction (+0.26pp). The gain cannot be attributed to retaining more tokens or using extra compute: held to flat's compute, the late schedule removes more tokens in total and leaves fewer tokens at the end, yet still wins. Single-layer probes point to a mechanism: earlier reductions perturb features that pass through more remaining layers, front-loading reduction error in depth. The effect is broad, holding across five token-reduction methods (ToMe, EViT, ATS, ATC, PiToMe), nine backbones, all ImageNet-C corruption types, eight further shift suites, and two further modalities, video and vision-language QA. It is also specific to shift, still positive on clean and rising monotonically to ~4x that at the highest severity 5. The schedule keeps its gain under six test-time adaptation methods, and needs no per-input or per-domain tuning.
Oct 6, 2026

Foveated Compression: Selective High-Resolution Preservation for Token-Efficient VLMs

Visual tokens are a major source of inference cost in vision-language models, yet simple image downsampling remains a surprisingly strong compression baseline. This raises a complementary question: under a fixed token budget, where should visual fidelity be preserved? We introduce Foveated Compression, which encodes a full-resolution image once and represents it with a mixture of native- and compressed-resolution visual tokens. A behaviorally self-distilled Foveated Merger compresses local visual tokens while preserving compatibility with their native counterparts, and a lightweight Foveated Selector chooses one of nine spatial cells to retain at native resolution using exhaustive budget-matched intervention supervision. At 11.11% visual tokens, uniform Foveated Compression shows no significant paired difference from iso-token downsampling. At 20.99%, the learned selector significantly outperforms random and fixed allocation, but remains below strong whole-image resizing, showing that localized fidelity is not universally preferable. A budget-matched region-choice oracle reaches 82.73 macro accuracy versus 69.61 for the learned selector, revealing substantial headroom within the same spatial action space. Matched probing further shows that signals predicting when compression breaks the answer are substantially more accessible after language-model computation than to the lightweight prefill-free selector. These results expose complementary bottlenecks in region selection and compressed-region fidelity.
Oct 4, 2026

When and What to Prune? Stage-Aware Visual Token Pruning for Efficient VLA

Visual token pruning is an effective way to accelerate vision-language models and is especially useful for vision-language-action (VLA) inference, where many visual tokens must be processed before predicting robot actions. Existing pruning methods usually estimate which tokens can be pruned based on attention scores or feature diversity, retaining tokens that are either highly attended or visually different from others. However, most of them use fixed pruning schedules, such as pruning once at a preset layer or pruning at uniformly spaced layers. Such schedules can be risky for VLA models, because the model may not know which visual regions matter for the action in early layers. Tokens that look unimportant at first may become useful after the model combines visual observations with the language instruction. In this work, we propose SAPrune, a training-free visual token pruning framework for efficient VLA inference. Instead of pruning at fixed layers, SAPrune uses a small calibration set to observe how action-to-visual attention changes across layers, and chooses pruning layers only after the attention pattern becomes more reliable. At each selected layer, SAPrune applies a dual-path pruning rule: one path protects strongly attended visual tokens from pruning, while the other prevents useful surrounding context from being discarded. Experiments on LIBERO, SIMPLER, and real-world robotic tasks show that SAPrune prunes 87.5% of visual tokens and achieves up to 1.718x inference speedup while maintaining competitive task success rates.
Oct 3, 2026

FlashGaze: Training-Free Multi-Scale Patch Pruning For Efficient Video Understanding

Multimodal Large Language Models (MLLMs) have demonstrated strong performance in video understanding, yet efficiently processing long, high-resolution videos remains challenging. Such videos often contain substantial spatiotemporal redundancy, and processing redundant visual tokens can incur avoidable computational overhead. Many existing methods prune visual tokens during or after vision transformer (ViT) encoding, leaving much of the encoding cost unaddressed. Some approaches prune patches before encoding but rely on learned auxiliary networks for patch selection, incurring additional training and inference overhead. To address these limitations, we propose FlashGaze, a training-free method that reduces spatiotemporal redundancy before ViT encoding without introducing auxiliary networks. FlashGaze uses pixel-space differences as a proxy for information loss and employs Quadtree Dynamic Programming to jointly optimize patch dropping, merging, and keeping under a fixed budget. Experiments on two MLLM backbones across multiple benchmarks demonstrate substantial efficiency gains while largely preserving accuracy. On Qwen3-VL-8B, FlashGaze retains 98% of the full-input baseline accuracy on LongVideoBench while achieving up to 5.4x and 17x speedups in ViT encoding and MLLM prefill, respectively, and reducing peak GPU memory usage by a factor of 1.8. These efficiency gains enable the model to process videos with more frames and higher resolutions on the same GPU hardware, unlocking video understanding at scales previously out of reach.
Oct 2, 2026

From Patching to Pruning Visual Computation in Vision Language Models

Vision language models (VLMs) incur substantial inference cost because every visual token is processed by the attention and MLP projections of every decoder layer, even when token-specific visual computation is unnecessary at many depths. We introduce Patch-to-Prune (P2P), inspired by Mechanistic Interpretability, a training-free framework that converts activation patching from a diagnostic tool into an inference-time computation bypass. P2P performs validation-guided forward and backward layer sweeps to identify decoder regions whose visual-token projection outputs can be replaced by fixed neutral proxy activation vectors within a user-specified accuracy tolerance. Unlike conventional token-pruning methods, P2P preserves the sequence length, token order, positional information, attention mask, and residual pathways, thereby pruning computation without removing tokens or modifying the pretrained model weights. We evaluate P2P on four VLMs from the Qwen2.5-VL and LLaVA families across seven multi-modal benchmarks using mutually disjoint calibration, validation, and test partitions. P2P at a 3% tolerance retains around 94% of dense accuracy while reducing FLOPs by 55%. Beyond these efficiency gains, our layer-wise analysis suggests that visual processing in VLMs is non-uniformly distributed across decoder depth: early and late layers often require little token-specific visual computation, whereas intermediate layers appear to perform most task-relevant visual integration, enabling later reasoning to rely largely on visual information already embedded in shared residual and textual representations. This makes P2P both an efficient inference framework and a causal lens into visual information processing in VLMs.
Sep 30, 2026

When Masking Helps or Hurts Robustness in Compressed CLIP: A Pre-Deployment Diagnostic

This paper demonstrate that whether masking-based token pruning helps or hurts worst-group robustness can be predicted before deployment, without labels or fine-tuning. A systematic study of semantic masking across 8 spurious-correlation benchmarks shows its effect on worst-group accuracy is highly unstable: it improves accuracy by up to 82.5% relative on some datasets and degrades it by up to 100% on others. We trace this instability to spurious inversion: background patches receive higher CLIP text-similarity than the true object when the spurious attribute is background-separable, inverting the assumption every text- and attention-guided pruning method relies on. We introduce the Spurious Inversion Metric (SIM), a label-free, pre-deployment diagnostic whose sign predicts this effect with statistical significance (binomial p=0.035p=0.035) across all 8 datasets, and remains dependable across 6 CLIP architectures with a clean foreground/background split. Naive masking is itself a major source of risk: it causes the largest average-accuracy loss of any method we evaluate, and its own per-image segmentation step is a significant runtime bottleneck. To address this, we design a batched, synchronization-free GPU segmentation routine that cuts this overhead from 3.5×\times to 1.75×\times baseline. Gating deployment by SIM's sign recovers masking's benefits while avoiding its worst failures, matching or exceeding a strong pruning baseline on 7 of 8 datasets.
Sep 30, 2026

DecoMoE: Decoupling Visual Propagation and Expert Computation for Efficient Multimodal MoE Inference

Multimodal mixture-of-experts (MoE) models combine sparse expert activation with visual-language capabilities, yet their inference remains costly because long visual-token sequences repeatedly incur attention, routing, dispatch, and expert-MLP computation. Existing methods typically compress either the token or expert dimension, leaving redundancy along the other. Our analysis reveals two complementary regularities: the depth required for visual propagation varies across inputs, while text-token routing exhibits concentrated and recurrent expert-importance patterns. Based on these observations, we propose DecoMoE, a two-dimensional structured compression framework that decouples visual propagation from expert computation. The Sample-Adaptive Visual Boundary (SAVB) predicts an input-dependent visual-exit layer at which the visual-token block is removed. The Routing-Calibrated Expert Prefix (RCEP) reorders experts offline using text-token routed mass and, from this predicted exit layer onward, retains at each MoE layer the shortest contiguous prefix covering a target routed-mass fraction. We evaluate DecoMoE on Qwen3-VL-MoE and InternVL3.5-30B-A3B across six benchmarks. On Qwen3-VL-MoE, DecoMoE retains 97.91% of dense-baseline performance while reducing computation from 27.06 to 16.73 TFLOPs and latency from 0.44 to 0.26 seconds, yielding a 1.69x speedup. Code will be available at https://github.com/ShawnTan86/DecoMoE.
Sep 29, 2026

TReVS: Integrating Textual Relevance and Visual Saliency for Efficient Vision-Language Model Token Pruning

Vision-Language Models (VLMs) excel at visual understanding and reasoning but often incur substantial inference costs due to the large number of visual tokens. Recent visual token pruning methods increasingly follow a two-stage paradigm: they first remove visually redundant tokens after the vision encoder and then discard tokens irrelevant to the textual query within the Large Language Model (LLM). However, since the first stage typically relies solely on vision-encoder saliency, it may prematurely eliminate query-relevant tokens, depriving the subsequent text-guided stage of critical visual evidence. Our empirical analysis shows that incorporating query guidance into first-stage pruning better preserves task-relevant evidence and consistently improves performance over vision-only saliency-based pruning. We further find that high-variance attention heads are more sensitive to the textual query and yield more discriminative text-to-vision attention signals for second-stage pruning. Motivated by these findings, we propose TReVS, a training-free framework that combines textual relevance with vision-encoder saliency for pre-LLM pruning and leverages high-variance attention heads to remove task-irrelevant tokens at shallow-to-intermediate layers of the LLM. On LLaVA-1.5-7B, TReVS retains 92.8% of the unpruned baseline performance while pruning 94.4% of visual tokens, outperforming prior state-of-the-art methods.
Sep 29, 2026

Beyond Token Importance: Preserving Spatial Scaffolds for Efficient Vision-Language-Action Inference

Existing VLA pruning strategies primarily select individual visual tokens according to task-level semantic relevance, while overlooking the spatial information required for robotic manipulation. To examine this limitation, we construct a simple Stride baseline that uniformly samples tokens along the flattened one-dimensional visual sequence, representing a purely geometric pruning strategy. Surprisingly, Stride outperforms semantic pruning and random pruning at certain pruning ratios, but collapses when the token budget is only slightly reduced. We characterize this phenomenon through the spatial coverage radius, defined as the largest spatial blind spot induced by the retained token set after pruning. Our analysis reveals a strong correlation between the spatial structure of retained tokens and task success, suggesting that reliable VLA pruning requires preserving not only task-relevant tokens but also the spatial scaffold of the scene. Motivated by this diagnosis, we propose GeoScaffold, a training-free visual token pruning method that partitions each image into spatial regions, allocates inter-region token budgets using task-relevance weights, and selects intra-region scaffold tokens via farthest point sampling to reduce the local coverage radius. On pi 0.5 and LIBERO, GeoScaffold retains only 20% of visual tokens while preserving a 93.2% average success rate, and achieves a 1.78 times prefill speedup over the unpruned baseline.
Sep 29, 2026

Representation Dynamics Reveal Semantic Saliency and Similarity for Visual Token Pruning in MLLMs

Multimodal large language models (MLLMs) incur high inference latency from long visual token sequences. Existing pruning methods commonly use attention maps or output features to estimate token importance or redundancy. Several recent approaches also exploit representation changes, but when and how these changes reflect foreground saliency and semantic consistency remain insufficiently understood. We analyze visual token representation dynamics across encoder depth and uncover two findings. First, the relationship between token update magnitudes and foreground saliency is layer-dependent: large token updates concentrate on foreground regions in two depth intervals, separated by several sink-dominated layers at intermediate depths. Second, similarities between token update directions better distinguish same-class from different-class tokens than those between encoder output features. Building on these findings, we propose MSDG-Prune, a training-free method that uses update magnitudes and directions to preserve salient and diverse visual information. Specifically, we group tokens by update-direction similarity and use query-weighted saliency derived from update magnitudes across a chosen depth window for group-wise token pruning. Extensive experiments across four MLLMs demonstrate the effectiveness and generalizability of MSDG-Prune. On LLaVA-NeXT, it retains 91.9% of uncompressed performance on average with only 5.6% of visual tokens, while achieving a 7.8x prefilling speedup. Code is available at https://github.com/liweixuan-hitsz/MSDG-Prune.
Sep 29, 2026

Decoding Affective Nuances: Enhancing MLLMs via Hierarchical Emotion Reasoning and Contrastive Discriminative Pruning

While multimodal large language models (MLLMs) have demonstrated exceptional capabilities in objective understanding tasks, their performance in affective reasoning still falls significantly short of human standards. We attribute it to a central capability gap: MLLMs are difficult to reliably distinguish semantically proximal emotions based on fine-grained visual evidence, which could be decoupled as two limitations: 1) Insufficient Attribution. The global reasoning paradigm of conventional MLLMs severely dilutes fine-grained emotion cues, where subtle emotional states are usually implicitly encoded, thereby generating emotional misjudgments in complex scenarios. 2) Insufficient Discrimination. Existing methods could only identify regions generally associated with emotions, which fails to distinguish discriminative regions between semantically similar emotions, leading to ambiguous emotion judgements. To overcome these limitations, we present a training-free inference-time optimization framework, named Decoding Affective Nuances (DAN). Specifically, we propose a Hierarchical Emotional Reasoning Chain (HERC) that enhances the insufficient attribution by harmonizing fine-grained scene/object-level cues and performing a soft-gated reasoning. Furthermore, to discriminate between semantically proximal emotions, we design a Contrastive Discriminative Visual Pruning (CDVP), which isolates discriminative visual tokens to reason the final emotion category by computing the absolute discrepancy between the attention distributions of similar emotions. Performances on several benchmarks demonstrate that DAN significantly improves discrimination for affective nuances without consuming additional training resources, especially achieving +10.47% improvements with Qwen3-VL-8B-Instruct on WebEmo25 dataset that contains 25 fine-grained emotion categories.
Sep 28, 2026

SPIDER: Multi-Layer Semantic Token Pruning and Adaptive Sub-Layer Skipping in Multimodal Large Language Models

Multimodal Large Language Models face significant efficiency challenges that stem from two distinct yet coupled sources: data redundancy and computational redundancy. While most methods focus on data redundancy by pruning visual tokens from the output of the visual encoder or computing redundancy in LLM decoders using blockwise importance, the finer-grained inter-layer representation shifts and the distribution differences within the layers themselves have not been fully explored. In this work, we comprehensively investigate this dual-level inefficiency. We posit that intermediate layer tokens from vision encoders should be considered for effective visual token pruning, as semantic focus shifts across layers, with middle-layer tokens capturing more detailed object-centric information that deeper layers may abstract away. Furthermore, we reveal the differential contributions of Attention and FFNs across distinct LLM decoder layers. Building upon these discoveries, we propose SPIDER, a training-free framework that integrates multi-layer Semantic visual token PrunIng with an aDaptive sub-layER skipping mechanism. Experimental evaluations demonstrate that SPIDER consistently maintains strong performance across various MLLM architectures and reduction ratios. For instance, on LLaVA-NeXT-7B, SPIDER reduces FLOPs by 79%79\% while maintaining 96%\% of the baseline performance.
Sep 28, 2026

P4Q: Co-designing Token Pruning and Quantization for Vision-Language Model Acceleration

Vision language models have achieved strong performance across a wide range of multimodal applications, yet their substantial computational and memory costs hinder efficient deployment. Visual token pruning and post-training quantization reduce inference overhead along two complementary dimensions, namely sequence length and numerical precision. Existing workflows typically optimize these techniques independently or apply them sequentially. Their distinct optimization objectives leave critical interactions unaddressed and constrain the achievable compression performance. We revisit these designs and present P4Q, a practical co-design framework that jointly optimizes visual token pruning and low-bit quantization for efficient VLM inference. First, P4Q introduces a quantization-aware visual token selection strategy before the LLM. It applies fake quantization to copies of the features produced by the projector and selects visual tokens using statistics computed from these fake-quantized features, thereby conditioning the selector's feature-based decisions on simulated low-bit perturbations. Second, P4Q introduces a pruning-aware quantization calibration strategy. It uses the same selection strategy as pruning to calibrate the quantized model on the retained-token distribution, thereby aligning the calibration process with the pruned execution path used during deployment. By coupling these two components, P4Q achieves substantial inference speedups while maintaining comparable task performance, resulting in a better efficiency-accuracy trade-off than independently optimized pipelines. For instance, on LLaVA-NeXT, P4Q achieves an average end-to-end inference speedup of 2.8x across eight distinct test sets, while retaining higher accuracy than prior compression and quantization methods.
Sep 28, 2026

When Text Matters: Design Principles for Visual Token Pruning in Vision-Language Model

Visual token pruning has been widely studied as a practical approach to reducing the computational cost of large vision-language models. However, it struggles to preserve essential visual information, which can lead to substantial performance degradation. In particular, image-based token selection can overlook task-relevant details, while text-guided token selection may fail to capture the text--visual relationships needed for complex reasoning. We find that applying textual guidance too early can limit its ability to identify answer-relevant visual regions, whereas text-to-visual attention becomes more informative at intermediate decoder depths. This finding motivates our training-free method, which separates early vision-guided pruning from deferred text-guided reselection. We first prune visual tokens using vision-encoder attention, retain additional candidates until the decoder midpoint, and then use text-to-visual attention to determine the final visual-token set. Across eight benchmarks and three models, our method outperforms the best-performing baselines by an average of 11.10 and 16.84 percentage points in performance recovery at 80% and 90% pruning, respectively, with comparable or lower LLM-prefill latency than most baselines. The source code is publicly available at https://github.com/kmc3661/DeFT
Sep 28, 2026

ACPruner: Visual Token Pruning as Biased Attention Coverage Maximization in LVLMs

Large Vision-Language Models (LVLMs) face significant computational inefficiencies caused by the large number of visual tokens. Existing visual token pruning methods mainly focus on either retaining individually important tokens or selecting mutually diverse ones. In this work, we revisit visual token pruning from a coverage perspective and formulate it as a biased attention coverage maximization problem. The key idea is to select a compact token subset whose encoder-side outgoing attention can jointly cover the image while assigning higher coverage priority to more informative regions. From this perspective, we propose ACPruner, a training-free visual token pruning framework for efficient LVLM inference. ACPruner first estimates token importance by combining intra-modal saliency and inter-modal relevance, then derives token-wise coverage from attention patterns within the vision encoder, and finally performs greedy selection to maximize the proposed coverage objective. Extensive experiments across multiple LVLM backbones, including LLaVA-1.5-7B/13B, LLaVA-NeXT-7B/13B, Qwen2.5-VL-7B, and LLaVA-OneVision-7B, show that ACPruner consistently achieves strong performance retention while delivering substantial end-to-end inference speedups.