VLM Inference

VLM: Vision-Language Model

Follow this topic Ask about this topic

Momentum

3 papers in the last four weeks, level with the four weeks before. 0.0% of all new papers.

Jul 13Week of Sep 28

Latest papers 56

Oct 7, 2026cs.CV

Why VLMs Miss Small Objects, and When Zooming In Is Safe

Vision-language models (VLMs) often miss small objects in large images. We ask three questions: what limits them, which of these limits better models can remove, and whether the classical way of handling large images, local decomposition, still has a future. We answer them with a theory built on two quantities of the image interface: S, the number of visual tokens across an object's side, and L, the content a call must cover. The limits: a W x H image sent whole within N tokens gives an object of side m at most m*sqrt(N/(WH)) tokens per side. Doubling the token budget N therefore raises the largest S a whole image can reach by only 41%, and seeing an image at the S an object needs costs at least order S^2 tokens whatever the model. If recognition improves gradually with S, any search strategy, zoom agents included, obeys a recall-cost frontier. What better models can change: the S an object needs and how much one call can carry, measured in bits per object found; the coverage cost remains. Decomposition: yes. Assuming only that more tokens per object and less content per call do not hurt on average, splitting an image cannot lower recall if no view zooms out relative to the whole image and views overlap by one object. Neither part of this condition can be dropped; we bound the cost of every such decomposition, and a simple rule approaches the bound as the image grows. We test the theory in about 177,000 requests on 797 images. On controlled images, none of 40 orderings predicted for 8 VLMs is violated. Checked after the fact on drawings, floor plans, natural and synthetic images, 61 of 89 implied orderings hold significantly and 4 fail, all for OpenAI models given more pixels than their default path. On construction drawings the rule never significantly lowered recall relative to the whole image and raised it by up to 0.28. Code and data: https://github.com/shijunzhe/vlm-small-objects

Junzhe Shi, Yuan Gan, Shida Jiang

Systems Engineering University of California, Berkeley Berkeley, CA, USA · Department of AI Quotr AI Emeryville, CA, USA

Vision-Language Models VLM Inference

Oct 5, 2026cs.AI

CVIF: A Criticality-Driven Visual Intervention Framework for Geometric Diagram Understanding in MLLMs

Despite significant progress in visual tasks by Multimodal Large Language Models (MLLMs), geometric diagram understanding remains challenging due to the presence of sparse visual cues and ambiguous symbol-primitive associations. MLLMs may therefore rely on textual priors, producing interpretations that conflict with visual evidence. We introduce the training-free Criticality-Driven Visual Intervention Framework (CVIF), an inference-time method that localizes critical layers and executes visual interventions during the transition from evidence aggregation to semantic decoding. At these layers, a Geometry-Constrained Local Relation Reconstruction (GCLR) module selects and weights vertex-centered visual evidence, while an Adaptive Visual Steering Operator (AVSO) redistributes attention mass toward the selected tokens. Experiments on PGPS9K and PGDP5K show that CVIF raises Overall F1 from 77.85 to 85.58 and from 75.23 to 82.84, respectively, establishing a novel inference-time visual intervention paradigm.

Jiahui Kang, Bifan Wei, Lingling Zhang +4

School of Computer Science and Technology, Xi’an Jiaotong University · Tencent Hy AI Data

Test-Time Adaptation of VLMs Visual Reasoning Visually Grounded Reasoning VLM Inference

Sep 29, 2026cs.CV

When to Retrieve, When to Stay: Uncertainty-Aware Temporal Evidence Allocation for Streaming Video-LLMs

Streaming video understanding requires Video Large Language Models (Video-LLMs) to reason over continuous visual streams under causal constraints. As the visual history grows, a bounded visual?processing budget requires evidence selection that balances temporal recency with query relevance. Recent-only selection excludes potentially relevant historical evidence, whereas Semantic-only retrieval can displace useful recent context when relevance scores are ambiguous. We introduce WRWS (When to Retrieve, When to Stay), a training-free framework for uncertainty-adaptive evidence allocation. A lightweight external vision-language encoder scores query relevance across the observed history, while an adaptive allocation module uses the normalized entropy of the similarity distribution as a proxy for retrieval uncertainty. WRWS favors semantic retrieval when relevance cues are reliable and strengthens the recency prior under uncertainty. Following a retrieve-first, encode-later pipeline, WRWS selects evidence before target-model visual encoding, such that only the selected observations are processed by the costly target Video-LLM. Experiments across four Video-LLM families and multiple model scales demonstrate competitive accuracy on StreamingBench and OVO-Bench. In our efficiency evaluation, WRWS reduces average vision-to-answer time to 47.93% of the state-of-the-art method. Code will be released.

Xiang Hu, Jiazuo Yu, Lu Zhang +2

IIAU Lab, Dalian University of Technology

Streaming Video Understanding Evidence Selection Query-Aware Video Frame Selection VLM Inference

Sep 28, 2026cs.CV

ReSight-SMC: Two-Stage Power Sampling via Island SMC with Visual Scouts

Power sampling has emerged as a training-free approach to LLM reasoning, eliciting capabilities comparable to reinforcement learning by sharpening the model distribution over complete responses. Despite this success, power sampling remains underexplored in large vision-language models (LVLMs). We transfer Power-SMC to LVLM decoding by defining a sequence-power target conditioned on both the image and the prompt. This direct transfer provides a strong training-free baseline, but leaves two aspects of finite-particle multimodal inference unaddressed. At the particle level, global resampling can collapse genealogies, while particle-based power sampling does not diversify trajectories through distinct visual cues in multimodal decoding, limiting exploration under a finite particle budget. At the answer level, sequence-level sharpening makes distinct reasoning trajectories compete even when they support the same answer. We introduce ReSight-SMC, a verifier-free two-stage power sampler for LVLM inference. Its first stage uses ancestry-isolated SMC islands to preserve independent trajectory families and routes a bounded set of prefix-conditioned visual scouts to prefix-relevant image regions while discouraging redundant overlap. Each scout temporarily increases attention to the image tokens and emphasizes its routed region. Exact importance correction preserves the base LVLM sequence-power target. The second stage aggregates terminal importance mass by canonical answer, powers the answer marginal, and samples an answer together with a supporting trajectory. Across four LVLM backbones and five benchmarks, ReSight-SMC achieves stronger aggregate performance than Power-SMC over both the reasoning and perception benchmark groups. Without post-training, it remains competitive in aggregate with backbone-matched models trained using reinforcement learning.

Yaowen Zhang, Xiangyu Qiu, Junyi Hu +5

Independent Researcher · University of Electronic Science and Technology of China · Tsinghua University

Markov Chain Monte Carlo Large Vision-Language Models VLM Reasoning VLM Inference

Sep 21, 2026cs.CV

INTCORT: Training-Free Spatial Reasoning Enhancement for Vision-Language Models via Input Transformations and Confidence Routing

Vision-Language Models (VLMs) have demonstrated remarkable capabilities in multimodal tasks, yet they still exhibit poor ability in spatial reasoning. Existing training-dependent and training-free enhancement methods suffer from high computational costs with catastrophic forgetting and internal mechanism interference that compromises general capabilities, respectively. In this work, we first verify two key hypotheses: appropriate geometric image transformation and query-reversal transformation can recover incorrect spatial predictions, and correct predictions exhibit higher relation-token confidence than incorrect ones. Based on these findings, we propose INTCORT, a training-free spatial reasoning enhancement framework that constructs multiple inference views through input transformations and aggregates their predictions via relation-token confidence routing, without modifying the VLM's internal mechanisms. Experimental results on several commonly-used benchmarks demonstrate that INTCORT substantially improves spatial reasoning accuracy across diverse VLMs, achieving an average improvement of 10.01% over all models and benchmarks. Compared with prior works, INTCORT achieves superior performance with improvements of up to 25.01%.

Haoran Sun, Jingqi Xu, Yanhui Li +3

The University of Hong Kong · University of Southern California · China Telecom +3

Visual Spatial Reasoning VLM Reasoning Test-Time Augmentation VLM Inference

Sep 1, 2026cs.CV

Compressing AI Traffic: Standardized Neural Network Coding of Visual-Token Representations in Split Vision-Language Inference

When the visual encoder and the language decoder of a vision-language model (VLM) run on different compute nodes, the intermediate visual-token embeddings become a communicated payload rather than an internal activation. We call such machine-consumed intermediate tensors AI traffic and ask how far they can be compressed with a standardized, training-free codec. We insert ISO/IEC 15938-17 Neural Network Coding (NNC) round trips on the complete visual interface of a Qwen3-VL-8B-Instruct video question answering pipeline, comprising the main visual-token representation and the DeepStack feature streams, while leaving weights, prompts, and generation untouched, and sweep the quantization parameter (QP) over a wide rate range. Closed-ended Video-MME accuracy remains close to the uncompressed reference up to a 98% reduction of the transmitted BF16 tensor and only then collapses; open-ended MLVU generation shows the same plateau-and-collapse profile under an LLM judge. This robustness is not due to near-lossless reconstruction: the decoded tensor is heavily discretized, carries substantial row-wise relative L2 error, and has a visibly steeper singular-value decay than its source. Downstream reasoning therefore depends on coarse structure and relative geometry rather than exact floating-point values, which argues for rate-task rather than rate-distortion optimization of AI traffic codecs.

Reza Heidari, Hamed R. Tavakoli, Juho Kannala

Aalto University, Nokia Technologies · Nokia Technologies · Espoo, Finland +1

Visual Token Compression VLM Inference

Aug 31, 2026cs.CV

LOCI: A Locator-Critic with Refinement Loop

Vision-Language Models (VLMs) still struggle on tasks requiring complex visual understanding. We argue that the core issue is not high-level reasoning, but instead failing to locate critical details in the image. Due to this shortcoming, VLMs generate often plausible but incorrect reasoning based on flawed perceptual grounding. To address this, we propose Locator-Critic (LOCI), a training-free framework that decouples visual search from evidence verification. LOCI employs a Locator agent to propose candidate visual evidence and a separate Critic agent to evaluate its relevance and sufficiency. These agents engage in an iterative refinement loop, progressively improving the evidence until it is adequate to answer the given question. This decoupled, self-correcting process yields substantial performance gains, achieving state-of-the-art results on multiple complex visual benchmarks. LOCI improves accuracy for both open-weight models like Qwen3-VL (+12.1 on V*, +5.8 on HR-Bench and +11.2 on VisualProbe-Hard) and proprietary models like Gemini 2.5 Pro (+8.9 on V*, +4.3 on HR-Bench, +4.8 on VisualProbe-Hard).

Walid Bousselham, Mathilde Caron, Arsha Nagrani +1

Google DeepMind

Visual Question Answering Vision-Language Grounding VLM Inference

Aug 21, 2026cs.CV

A VLM Answer Is Not an Anomaly Score: Rank Compression Across Image and Video Anomaly Detection

Anomaly detection aims to identify observations that deviate from normal patterns. Recent work uses pretrained vision-language models (VLMs) for training-free image and video anomaly detection without task-specific retraining. Anomaly detection is commonly evaluated by how well anomaly scores rank anomalous images or video frames above normal ones. Generative VLMs, however, assign probabilities to possible answers and then decode a single answer. This decoding step can discard ordering information. We call this loss of ordering decoded-answer rank compression and study whether it materially affects anomaly detection performance. To isolate this effect, we compare two ways of scoring the same VLM output: one uses only the decoded answer, while the other computes a probability-weighted score over all possible answers. Across image and video anomaly detection benchmarks, VLMs, and answer scales, probability-weighted scoring consistently outperforms decoded-answer scoring, with mean gains ranging from 7.66 to 19.95 points on the primary benchmark metrics. Using answer probabilities only to break ties created by decoded-answer scoring recovers at least 95% of the average performance gap on every benchmark. When answer probabilities are available, how VLM answers are converted into anomaly scores is therefore part of the detector design, not merely an implementation detail.

Inpyo Song, Jangwon Lee

SungKyunKwan University South Korea

Video Anomaly Detection Zero-Shot Anomaly Detection Anomaly Detection VLM Inference

Aug 8, 2026cs.CV

Your VLM Already Knows When: Training-Free Temporal Grounding by Asking Yes or No

Multimodal LLMs that recognise events reliably still fail to say when they happen. Prompted for timestamps, strong VLMs reach as little as

3.8\%

[email protected] on Charades-STA, and

77

to

80\%

of their wrong predictions carry low output entropy: the models are confidently wrong, and entropy-based error detection stays below a random classifier. We show that this failure lives in the task interface, not in perception. Holding the weights fixed, replacing timestamp regression with a coarse-to-fine scan of binary questions, whose first-token probabilities are consumed only as a ranking, raises [email protected] by

28

to

50

points across four frozen backbones. The residual failures decompose into two measurable axes: a perception axis that moves with the backbone, and a geometry axis that is analytically predictable from the ratio of the output-window and event widths. FV-Action, the training-free method built on this analysis, reaches

56.8\%

[email protected] on Charades-STA, above the same backbone's native grounding pipeline and the strongest training-free result on this benchmark; it surpasses every TVG-trained model evaluated zero-shot on TACoS, and improves over direct prediction on ActivityNet Captions and QVHighlights, with no temporal supervision at any stage.

Ji Huang, Barry Devereux, Hui Wang

Queen’s University Belfast

Temporal Video Grounding Temporal Reasoning in Language Models Temporal Action Localization VLM Inference

Aug 8, 2026cs.CV

How Much Does It Cost to Answer My Question? Benchmarking Cloud VLM-based VQA Systems

Vision-language models (VLMs) are becoming a practical backend for mobile visual question answering (VQA) systems, enabling smartphones and smart glasses to answer users' questions about the physical world. Since modern VLMs remain difficult to run on mobile and edge devices, VQA systems increasingly offload inference to cloud-based VLMs. This gives mobile devices access to stronger computation, but it also makes visual input preparation a key system variable: how the image is prepared before offloading affects not only answer quality but also payload size, token cost, and system latency. Proprietary APIs expose little control over model internals or serving behavior, leaving client-side preprocessing as the main practical optimization space for downstream developers. Many such techniques have been proposed for visual offloading, yet their cost-quality impact on commercial cloud VLMs has never been studied. To fill this gap, we present VQABench, the first systematic benchmark that treats client-side input preprocessing as a controlled variable for cloud-VLM-based VQA. We evaluate 12 preprocessing techniques across three VQA datasets and four commercial VLMs from three providers, totaling 95,168 API calls. Our results show that preprocessing is not universally beneficial: its effectiveness depends on the target model, API paradigm, provider token-accounting rule, and task formulation. A poorly selected preprocessing strategy can increase deployment cost or latency while degrading answer accuracy. Overall, our benchmark clarifies when preprocessing helps, when it fails, and why, providing insights to guide future research and real-world deployment of VQA systems.

Henri Vanhuynegem, Weitao Xu, Yiran Shen +1

Department of Software Technology, Delft University of Technology, Delft, The Netherlands · Department of Computer Science, City University of Hong Kong, Hong Kong SAR, China · School of Software, Shandong University, Jinan, China +1

VLM Evaluation Visual Question Answering Efficient VLM Inference VLM Inference

Aug 6, 2026cs.CV

SCI-CLIP: Segment-Centric Inference with Reference Memory for Training-Free Open-Vocabulary Segmentation

Training-free open-vocabulary segmentation remains limited by a missing inference abstraction. Frozen vision-language features are produced at patch level, yet dense prediction requires a unit that simultaneously governs feature interaction, spatial support, contextual recovery, and retrieval-based correction. We present SCI-CLIP, a segment-centric inference framework built around the principle that the same region abstraction should organize all stages of dense open-vocabulary prediction. SCI-CLIP first induces a region-consistent interaction graph over frozen visual tokens, then reconstructs dense features by propagating values over this graph, augmenting them with selective cross-window support only where local evidence is insufficient. The same segment abstraction is subsequently used to construct and query an offline reference memory, aligning exemplar retrieval with the units on which prediction is made. SCI-CLIP turns frozen CLIP-style features into spatially coherent, context-aware, and retrieval-compatible dense predictions without any training. SCI-CLIP consistently improves the structural quality of dense predictions, the robustness of contextual reasoning, and the alignment of exemplar-based correction, yielding stronger open-vocabulary segmentation across eight benchmarks. Project code is available at: https://github.com/mzamini92/SCICLIP.

Mohamad Zamini, Diksha Shukla

University of Wyoming Laramie, WY

Memory-Augmented VLMs Open-Vocabulary Semantic Segmentation VLM Inference

Jul 25, 2026cs.CV

Similarity Is Not Logic: Factored Inference for Dual-Encoder Vision-Language Models

Dual-encoder vision-language models (VLMs) expose a similarity interface that enables zero-shot retrieval but fails compositional constraints: queries like "umbrella and no person" retrieve images containing both, even when concept detection is reliable. We trace this to an interface-level Bag-of-Concepts effect, where similarity scores approximate mean pooling of concept evidence regardless of operators. Although operator-dependent signals exist in text embeddings, they are too weak or misaligned to affect rankings. Fine-tuning does not reliably resolve this failure because the dominant bottleneck is how similarity aggregates evidence rather than what encoders represent. We propose factored inference, which separates evidence extraction from constraint execution, and introduce LCSE (Logic-Constrained Score Editing), a training-free method that executes constraints externally using concept scores from frozen encoders. We also introduce FACTOR-Bench, where LCSE achieves 85.5% accuracy versus 73.2% for the best fine-tuned baseline, 90.7% when applied to SigLIP 2, and improves NegBench COCO MCQ accuracy from 27.2% to 65.2% while preserving retrieval performance.

Sultan Alshehri, Zhantao Yang, Han Zhang +1

Carnegie Mellon University, Pittsburgh, PA, USA.

Vision-Language Models VLM Reasoning Vision-Language Retrieval VLM Inference

Jul 16, 2026cond-mat.other

A Modern Multimodal Assistant on a 6 GB 2011 GPU: Stage-Validated, All-GPU CUDA Inference for Fermi

A companion study ran a 35B mixture-of-experts model on a 2011 NVIDIA Tesla C2075 (Fermi, sm_20, 6GB) as a GPU-prefill/CPU-decode hybrid, because the 4-bit model did not fit in device memory (arXiv:2606.24031). This report keeps the hardware and asks what a model that fits can do: we deploy MiniCPM-V-4.6, a modern multimodal assistant pairing a SigLIP2 vision encoder and window-attention merger (16x visual token compression) with a compact hybrid gated-delta-net backbone, entirely on the GPU. Three results. (i) An all-GPU engine built on measured foundations: projections that dequantize 8-bit weights once and call the vendor SGEMM still in the last Fermi toolchain (64% of FP32 peak; our best hand-written GEMM hit 37%, wrongly called the ceiling); a chunked delta-rule rewrite of the recurrent layers, 2.8x faster than the sequential scan once attribution exposed one bad kernel; and a measured negative: 4-bit weights make decode slower than 8-bit here, since Fermi issues nibble-unpacking shifts at half rate. (ii) The vision side is a port with a proof obligation: we translate tower, merger, and projector to sm_20 CUDA, validating every stage against a locally generated reference forward (full tower 1.4e-5). One failure, position-embedding bucketization differing on exact rational ties, generalizes to a rule: float tie-breaking in index arithmetic is implementation-defined; call the reference operator, do not reimplement it. (iii) Long context exposes an O(N^2) wall short benchmarks hide: prefill falls from 114 tok/s at 2k tokens to 21 at 10k in a naive attention kernel; per-head vendor-GEMM calls writing into the existing score buffer (zero extra memory) restore a flat profile (408 at 2k, 361 at 10k; 17x), verified by exact needle retrieval from 60% depth. The same rewrite cuts image encoding 6x, to 0.93s. The system answers an image question end-to-end in 1.7s.

A. C. Opus, J. Q. Lu

Department of Physics, University of Puerto Rico, Mayaguez, PR 00680, USA

Efficient Transformer Inference GPU Kernel Optimization Efficient VLM Inference Multimodal Foundation Models VLM Inference

Jul 14, 2026cs.AI

What Keeps Vision-Language Models Looking at the Image?

When do vision-language models need direct access to the image while generating an answer? We study image dependence during answer generation by examining how the visual information needed for the current question becomes available in context. We intervene on direct image access while retaining previously computed states. Across real-image and synthetic tasks, we show that, depending on the generation process, direct access can continue to support accuracy after question processing. Supplying the required attributes as text in the context weakens this dependence. On synthetic tasks, we also examine how dependence changes as the model itself states the required attributes. Before attribute expression, severing access reduces accuracy, and replacing image-side states shifts answers toward the counterfactual content. After sufficient expression, both interventions have smaller effects. Even with an identical generated prefix, dependence differs according to whether image access was available during question processing. Thus, both the visible text and the preceding image access matter. Several of these patterns hold across model families, including Qwen2.5-VL-32B and InternVL3-14B. These findings offer a view of image dependence in terms of the information needed for the current question and the history of image access, beyond generation position alone. This perspective provides a basis for deciding when to reduce visual access during an answer and which visual information to retain for subsequent questions.

Hiroto Osaka, Shohei Taniguchi, Gouki Minegishi +3

The University of Tokyo

Vision-Language Models VLM Interpretability VLM Inference

Jul 8, 2026cs.CV

Sparse Attention for Dense Open-Vocabulary Prediction in CLIP

Contrastive Language-Image Pre-training (CLIP) relies on softmax-based self-attention, a strictly positive distribution that assigns probability mass to every pair of tokens-even semantically irrelevant ones. While these dense softmax weights are effective for gathering broad context during pre-training, they spread attention across many low-salience tokens, producing noise that obscures the fine-grained, spatially localized cues required for dense, open-vocabulary prediction. We study an inference-time substitution of the row-wise softmax in the final visual self-attention layers with the

α

-entmax transform, applied across both the standard query-key attention and self-correlation variants. Because entmax applies a data-dependent threshold that maps low scores exactly to zero, it acts as an implicit denoiser, zeroing contextually irrelevant dependencies while redistributing mass onto the most relevant tokens. We evaluate on open-vocabulary tasks-dense semantic segmentation (Pascal VOC, Pascal Context, ADE20K) and fine-grained retrieval (FG-OVD)-and find the gain from attention sparsification is proportional to how much the baseline attention spreads off the target class.

Fatimah Zohra, Chen Zhao, Shuming Liu +1

King Abdullah University of Science and Technology (KAUST) · Harbin Institute of Technology, Shenzhen

Visual Attention Fine-Grained Image Retrieval Sparse Attention Open-Vocabulary Semantic Segmentation VLM Inference

Jul 6, 2026cs.CV

QSVideo: Query-Conditioned Semantic Temporal Retrieval for Video Understanding

The performance of vision-language models (VLMs) in video understanding declines with increasing video duration, as video moments unrelated to the query confuse their language components. Multimodal retrieval has emerged as a critical component of video understanding, addressing this challenge by localizing key visual evidence. However, existing multimodal retrieval methods suffer from biased relevance estimation, limited diversity, and temporal collapse. In this paper, we propose QSVideo, a unified framework that systematically addresses relevance, diversity, and temporal modeling in video retrieval. We first introduce a query-conditioned semantic ranker, QSRanker, which reformulates arbitrary questions into retrieval-friendly queries and estimates structured relevance along object, action, and location dimensions. Building upon this, we design QSRetrieval to jointly optimize relevance and diversity for more informative frame selection. Moreover, we propose temporal alignment strategies tailored for both long and streaming videos to improve evidence recall. Extensive experiments on long and streaming video benchmarks demonstrate that QSVideo greatly enhances video VLM performance under strict frame limit constraints. The code is available at https://github.com/human-analysis/QSVideo.

Wei Ao, Lan Wang, Vishnu Naresh Boddeti

Michigan State University, East Lansing MI 48824, USA

Temporal Video Grounding Streaming Video Understanding Long-Video Understanding Video Retrieval Query-Aware Video Frame Selection VLM Inference

Jul 2, 2026cs.CR

Vision Token Manipulation Attacks on Cloud-Edge Inference of Large Vision-Language Models

Cloud-edge Large Vision-Language Model (LVLM) inference enables efficient deployment by splitting computation between edge devices and cloud servers. In this process, intermediate vision tokens are transmitted from the edge to the cloud over a communication link, thereby exposing a new attack surface. We study vision token manipulation attack (VTM-Attack) under a black-box man-in-the-middle setting, where an adversary intercepts and manipulates a subset of transmitted vision tokens under a budget constraint. We propose four naïve attack strategies and an optimization-based token selection method. Experiments on 6 state-of-the-art LVLMs (3B-72B) across 4 benchmarks show that manipulating only 10% of vision tokens can reduce accuracy by up to 88.31%. These results reveal a critical vulnerability in cloud-edge LVLM inference.

Zikai Zhang, Rui Hu, Olivera Kotevska +1

Department of Computer Science and Engineering, University of Nevada, Reno, Reno, USA · Oak Ridge National Laboratory, Oak Ridge, USA

Adversarial Attacks on VLMs Large Vision-Language Models VLM Inference

Jul 2, 2026cs.CV

ProCal: Inference-Time Proposal Calibration for Open-Vocabulary Object Detection

Open-vocabulary object detection aims to localize and classify objects beyond the fixed set of categories seen dur ing training. Recent open-vocabulary object detection methods improve localization and classification for unseen categories by leveraging a frozen VLM as a detector backbone. However, VLM classification score lacks recognizing position and scale of the object in an image. We observe that pretrained VLMs en able to classify foreground and background regions. According to this observation, we propose a simple inference-time Pro posal Calibration (ProCal) that improves localization quality of the classification score. ProCal computes a proposal prior by combining two scores: localization-aware foreground score and background-aware suppression score. Localization-aware foreground score captures whether a proposal contains an object area. Background-aware suppression score measures the extent to which the proposal resembles background. We analyze that ProCal suppresses false novel activation on background proposals and consistently ranks true novel proposals above background and partial novel proposals. Applied to CLIPSelf ViT-L/14, ProCal improves APr +2.5 on OV-LVIS. The analyses show that proposal-level localization-aware reranking effects to mitigate ranking miscalibration for novel objects.

Jae-Ryung Hong, Ho-Joong Kim, Seong-Whan Lee

Department of Artificial Intelligence, Korea University, Anam-dong, Seongbuk-ku, Seoul 02841, Korea

Open-Vocabulary Object Detection VLM Inference

Jul 1, 2026cs.CV

Towards High-Resolution Visual Perception via Hierarchical Entity Exploration

High-resolution (HR) image perception remains a key challenge in multimodal large language models (MLLMs), as fine-grained details are often lost when the image is processed as a whole. Existing methods either require training to teach models where to look or heuristically divide the image into fixed regions, both of which struggle to generalize in complex HR scenes. In this work, we propose Hierarchical Entity Exploration (HEE), a training-free and model-agnostic framework that transforms static image understanding into dynamic, query-guided entity exploration. HEE first evaluates each region using a dual scoring mechanism to determine whether it already contains sufficient evidence to answer the question. If not, it applies object detection within the most promising region to extract fine-grained entities, clusters them into coherent subregions, and organizes them into a multi-level semantic hierarchy for deeper exploration. When deeper regions still fail to yield confident answers, a confidence-guided backtracking mechanism revisits alternative paths to ensure adaptive perception. Extensive results show that HEE outperforms training-free methods like ZoomEye and RAP in both accuracy and efficiency on two complex HR benchmarks (Visual Probe and HR-Bench), across different MLLMs such as Qwen2.5-VL and LLaVA-OneVision. Moreover, HEE demonstrates generalization on the MME-RealWorld benchmark.

Ziyu Ma, Shidong Yang, Yuxiang Ji +5

AMAP, Alibaba Group, China · Monash University, Australia

Large Vision-Language Models Fine-Grained Visual Perception Multimodal Understanding VLM Inference

Jun 25, 2026cs.CV

Dismantling Pathological Shortcuts: A Causal Framework for Faithful LVLM Decoding

Large Vision-Language Models (LVLMs) exhibit sophisticated reasoning but remain susceptible to object hallucination. Deviating from the prevailing attention intensity assumption, we reveal a deeper dynamic structural misalignment: hallucination is triggered at decision-critical steps where specific attention heads, acting as risky mediators, decouple from visual evidence to lock onto language priors. This establishes a pathological shortcut that bypasses visual grounding. To dismantle this, we propose Fox (Faithfulness and Observational-flow via eXpression-rectification), a training-free inference-time framework. Fox diagnoses structural misalignment using a visual attention entropy probe to localize risky mediators unsupervisedly. We then execute a targeted causal intervention via numerical logit saturation to physically sever the shortcut path. Finally, a conflict-gated cooperative decoding strategy reconciles interventional faithfulness with observational fluency. Extensive experiments demonstrate that Fox achieves SOTA performance, outperforming SID by 29.1% while preserving linguistic richness. Code is available at https://github.com/Cc2021start/Fox.

Liu Yu, Can Chen, Ping Kuang +3

School of Information and Software Engineering, University of Electronic Science and Technology of China, Chengdu, Sichuan, China · School of Computer Science, University of Auckland, Auckland, New Zealand

Object Hallucination in VLMs Faithfulness in VLMs VLM Hallucination Mitigation VLM Inference

Jun 16, 2026cs.CV

MLLMs Get It Right, Then Get It Wrong: Tracing and Correcting Late-Layer Textual Bias

When vision contradicts text, multimodal large language models (MLLMs) consistently favor text, even when images provide clear evidence otherwise. This bias poses risks for applications requiring visual grounding, yet its cause remains unclear. In this paper, we uncover a surprising finding: models often get it right initially, forming correct vision-based predictions in their intermediate layers, before changing their minds and favoring text in the final output. We call this "late-layer textual override". The visual information is encoded, it simply does not survive to the output. More intriguingly, we find that how predictions change reveals whether they're correct: 85% of failures shift toward text, while 89% of successes shift toward vision. This directional signature enables a simple but powerful intervention: when we detect a confident visual prediction being suppressed, we restore it. We propose CALRD (Conflict-Aware Layer Reference Decoding), a training-free method that recovers overridden predictions at inference time. Experiments across five MLLMs of varying architectures demonstrate up to 9.4% absolute improvements on conflict benchmarks while largely preserving standard performance, without training or external knowledge. It recovers what the model already knew but failed to preserve.

Xingming Li, Ao Cheng, Qiyao Sun +4

National University of Defense Technology, changsha, China · 2Chinese University of Hong Kong, Shenzhen, China · 3Intelligent Game and Decision Lab, Beijing, China

VLM Robustness Vision-Language Models Vision-Language Grounding Modality Gap in VLMs Visually Grounded Reasoning Multimodal Foundation Models+2

Jun 16, 2026cs.CV

LiveStarPro: Proactive Streaming Video Understanding with Hierarchical Memory for Long-Horizon Streams

Despite the remarkable progress of Video Large Language Models (Video-LLMs), current online architectures still struggle to simultaneously process continuous video streams, decide autonomously when to respond, and preserve long-horizon contextual memory. These obstacles undermine real-time responsiveness and cause severe forgetting throughout prolonged interactions. In this work, we introduce LiveStarPro, a live streaming assistant that is designed for proactive video understanding over long-horizon streams. The design of LiveStarPro rests on three complementary components. The first component is Streaming Verification Decoding (SVeD), an inference framework that identifies the appropriate response timing through single-pass perplexity verification, thereby eliminating the dependency on explicit silence tokens. The second component is Streaming Causal Attention Masks (SCAM), a training strategy that enforces incremental video-language alignment over variable-length streams. The third component is Tree-Structured Hierarchical Memory (TSHM), a recursive memory architecture that organizes evicted historical information into event chains and consequently enables efficient retrieval from effectively unbounded video streams. To facilitate a comprehensive evaluation under realistic online conditions, we further present OmniStarPro, a large-scale benchmark that spans 15 diverse real-world scenarios and that extends to hour-scale streams for the assessment of long-term recall. Extensive experiments demonstrate that LiveStarPro consistently surpasses existing methods, attaining a 28.9% improvement in semantic correctness and an 18.2% reduction in timing error, while its streaming key-value cache further yields a 1.58x inference speedup over the same model without caching. The model and the code are publicly available at https://github.com/sotayang/LiveStarPro.

Zhenyu Yang, Kairui Zhang, Bing Wang +2

Streaming Video Understanding Long-Video Understanding Hierarchical Memory VLM Inference

Jun 15, 2026cs.CV

VisualClaw: A Real-Time, Personalized Agent for the Physical World

Vision language models are serving as general-purpose interfaces for complex multimodal tasks. However, deployment still faces three gaps: VLMs typically incur high latency and cost when processing dense video frames and long prompts, the agent scaffold remains static after deployment, and standard video-QA benchmarks do not test whether agents can use visual evidence inside tool-using workspaces. We present VisualClaw, a self-evolving multimodal agent built around two principles. First, hybrid encoding reduces deployment cost by filtering less informative streaming frames with a cascaded gate and compressing the text skill bank through hot/cold top-k injection. Second, skill evolution lets the agent learn from failures: retrieved memories condition an evolver as direct concatenated context or as guided evidence, producing skill-bank updates that help future questions. Across 4 video-QA benchmarks with 2 VLMs, VisualClaw cuts per-question API cost by an average -98% versus full-frame upload and by -25.9% over the offline uniform 8 frame baseline, while boosting accuracy in most settings, e.g., an average +3.85% and a peak +15.80% on EgoSchema with Gemini 3 Flash. To address the gap, we curate VisualClawArena, a 200-scenario multimodal agentic benchmark built through a strict five-stage pipeline; models must use video evidence, documents, dynamic updates, and executable checks inside a workspace. On VisualClawArena, the same framework with computer-use agent backends improves macro accuracy by +2.9% for Codex (GPT-5.5) and +3.2% for Claude Code (Sonnet 4.6) over no-evolution baselines, with a -9.5% cost reduction compared to the uniform-sampled baseline. These properties make VisualClaw a natural fit for edge applications, where the cascade reduces a 1-hour streaming session from ~3,600 API uploads down to only 5-20 calls and the self-evolution makes it a perfect personalized assistant.

Haoqin Tu, Jianwen Chen, Zijun Wang +14

1UC Santa Cruz · 2UNC-Chapel Hill · 3Google +1

Continual Learning for LLM Agents AI Agent Benchmarks Tool-Using Vision-Language Agents Memory-Augmented Language Model Agents VLM Inference

Jun 5, 2026cs.CV

Don't Pause: Streaming Video-Language Synchrony for Online Video Understanding

Online Video Large Language Models (Video-LLMs) have advanced toward seamless human-AI interaction through frame-by-frame processing and proactive responding. However, a critical challenge remains in streaming scenarios: existing models typically pause video perception while generating responses, breaking real-time video-language synchrony and causing stutters. To address this, we introduce a novel paradigm for online video understanding: Streaming Video-Language Synchrony (SVLS), and present LyraV, a live streaming assistant built upon a hierarchical control framework with two core innovations. First, the Frame-Driven Transition Controller (FDTC), a training-free verification-based finite-state machine, makes high-level semantic decisions on when to continue speaking, start a new response, or stay silent. Second, the Streaming Token Pacer (SToP), a plug-and-play lightweight predictive module, dynamically adapts the language generation rate to match the pace of the visual content. Concretely, LyraV performs \emph{per-frame incremental, sub-budget decoding}: within each frame interval it emits only a small chunk of tokens that fits the real-time budget, so perception is never blocked for a full sentence. Together, these components enable LyraV to seamlessly interleave incoming video frames with generated word tokens, achieving a fine-grained synchrony. Extensive experiments conducted on five online and three offline benchmarks demonstrate that LyraV preserves the backbone's general understanding ability while substantially improving streaming synchrony and narrative fluency, delivering a 98.29% synchrony with video playback and a real-time processing speed of 3.89 FPS. Interestingly, we observe an empirical capability in LyraV: dynamic reasoning over streaming tokens, enabling continuous interpretation and "thinking" alongside visual input.

Zhenyu Yang, Kairui Zhang, Shengsheng Qian +2

Streaming Video Understanding Language Model Decoding VLM Inference

Jun 5, 2026cs.AI

NAVI-Orbital: First In-Orbit Demonstration of a Zero-Shot Vision-Language Model for Autonomous Earth Observation

As Earth Observation data generation outpaces downlink bandwidth and human-in-the-loop processing, a widening gap has emerged between onboard collection and actionable ground intelligence. This paper presents NAVI-Orbital, a software system deployed on a Low Earth Orbit (LEO) spacecraft. On April 16, 2026, NAVI-Orbital achieved what is, to the authors' knowledge, the first in-orbit demonstration of a vision-language model performing autonomous multi-modal inference entirely onboard. NAVI-Orbital uses a local vision-language model (Gemma 3) to classify each captured scene, produce a text description of its content and the relationships between its features, and respond to operator follow-up via natural-language dialogue. The system is re-tasked through plain-English prompts in place of conventional command sequences, and is orchestrated by a graph-based state machine (LangGraph) coordinating dedicated agents for detection and dialogue. Results across ground benchmarking (88.16% accuracy on the 7,960-image curated AID benchmark), Flatsat validation, and live in-orbit captures of newly acquired, previously unseen Earth imagery (including uncorrected YAM-9 imagery, processed onboard with hardware-accelerated GPU inference and no fine-tuning for the flight instrument) demonstrate the feasibility of running foundation models on satellite-class edge computers to invert the conventional acquire-then-downlink-everything bandwidth profile through semantic compression of Earth observations in-orbit.

Juan Manuel Delfa Victoria, Taran Cyriac John, Andrew W. Herson

Jet Propulsion Laboratory California Institute of Technology Pasadena, CA, USA · School of Engineering and Computer Science Victoria University of Wellington Wellington 6140, New Zealand · Loft Orbital 321 11th St, San Francisco, CA 94103

Remote Sensing Image Understanding Vision-Language Models VLM Inference

May 31, 2026cs.CV

Disagreement-Based Cross-Model Routing for Implicit Video Question Answering

We study multiple-choice video question answering on the ImplicitQA benchmark, where the correct answer is never explicitly shown but must be inferred from off-screen events, line-of-sight cues, causal structure, and cross-shot spatial layout. On this benchmark a single frontier video LLM already operates near its accuracy ceiling, and we observe that conventional self-consistency strategies -- majority voting across repeated samples of the same model -- can hurt rather than help, because the model's errors on hard questions are correlated. We propose disagreement-based cross-model routing, a pure inference-time procedure that requires no labels and no training. We triple-sample a native-video model (Gemini 3.1 Pro Preview) at temperature zero, exploit the genuine sample-to-sample variance of its video-processing pipeline to identify the roughly 20% subset of questions where the three samples disagree, and route only that subset to a second model from a different family (Claude Opus 4.8) that consumes uniformly sampled frames with adaptive thinking. On the 1001-question validation set with public ground truth -- our main evaluation -- the method improves AvgAcc by +1.43 over the best single sample of the primary model, with per-category gains concentrated on Motion & Trajectory (+5.49), Inferred Counting (+3.45), and Vertical Spatial Reasoning (+1.82) -- the categories most dependent on cross-shot reference resolution. The same pipeline applied to the held-out 172-question CVPR 2026 ImplicitQA challenge test set achieves 82.03 AvgAcc / 79.71 MacroAvgAcc (+1.81 over the best single sample of the primary model), confirming the validation result on an independent split.

Durga Sandeep Saluru

Independent Researcher

Question Answering Video QA VLM Inference

May 30, 2026cs.CV

The Right Inference Strategy Is All You Need: Nearly Training-Free Domain-Wise Inference for EgoCross Challenge

EgoCross evaluates multimodal large language models on egocentric video question answering under substantial domain shift, where test videos come from surgery, industrial assembly, extreme sports, and animal-mounted cameras rather than ordinary daily-life scenes. In the source-limited track, the base model is fixed to Qwen3-VL-4B, while the official task-specific support set contains only 20 training samples. This setting makes the challenge less about model scaling and more about exposing the right visual, temporal, and answer-selection cues to a constrained model. Our key observation is that the frozen baseline model is not simply incapable of these rare scenarios; rather, it often fails to transfer its existing visual-language knowledge to the new task format without an appropriate interface. We therefore use a domain-wise inference strategy that treats the four target domains separately and designs different input, prompting, and answer-mapping procedures according to each domain's task characteristics. These strategies make the rare egocentric scenes more interpretable to the VLM by emphasizing the cues that matter for each domain. The resulting system is nearly training-free: surgery, and animal questions are answered with the base Qwen3-VL-4B model, while XSports and industry use only the official SFT checkpoint trained for two epochs on the provided 20 training samples. On the final evaluation, this simple strategy reaches 66.98% overall accuracy, suggesting that careful domain-aware inference can compensate for limited base-model strength and recover much of the ability already present in the baseline model.

Leyi Wu, Yifan Zhao, Jinjie Zhang +2

HKUST(GZ) · Knowin · HKUST

Egocentric Video QA Video QA Egocentric Video Understanding Cross-Domain Transfer Learning Multimodal Understanding VLM Inference

May 29, 2026cs.LG

Diversity Matters: Revisiting Test-Time Compute in Vision-Language Models

Test-time compute (TTC) strategies have emerged as a lightweight approach to boost reasoning in large language models (LLMs). However, their application and benefits for vision-language models (VLMs) remain underexplored. We present a systematic study of TTC across seven VLMs and six benchmarks, specifically analyzing feature-based scoring and majority voting methods. We find that feature heuristics fail and voting yields only modest gains in single-model settings. We theoretically show that this limitation stems from a lack of prediction diversity: when outputs are highly correlated, voting provides little benefit. In contrast, multi-model ensembles offer richer diversity, yet standard majority voting fails to account for varying model capabilities. To address this, we propose Entropy-based TTC (ETTC), which selects the most confident prediction based on predictive entropy. Our method reduces to majority voting in the single-model case, but in model ensembles, it leverages confidence disparities to prioritize stronger models. We prove that ETTC outperforms majority voting under mild assumptions and empirically demonstrate that it consistently surpasses both voting and the best individual model. Crucially, our results show that smaller models can synergistically enhance larger ones, unlocking ensembling gains not achievable with standard strategies.

Yijie Tong, Yifan Hou, Shaobo Cui +2

ETH Zürich · Shanghai Jiao Tong University · EPFL

VLM Evaluation Vision-Language Models Selective Prediction Test-Time Scaling for VLMs Ensemble Diversity VLM Inference

May 28, 2026cs.CV

Mitigating Hallucination in Vision-Language Models through Barrier-Regulated Adaptive Closed-form Steering

Large vision-language models (LVLMs) often hallucinate objects that are not present in the input image, largely because visual grounding weakens as decoding progresses. Existing inference-time mitigation methods modify logits or hidden states throughout generation, but they suffer from three key limitations: they lack an explicit grounding objective, intervene even when the model is already well-grounded, and use fixed correction strengths that do not adapt to the severity of grounding failure. We propose BRACS (Barrier-Regulated Adaptive Closed-form Steering), a training-free steering framework that addresses these issues through barrier-regulated adaptive closed-form steering. BRACS monitors the model's own attention to measure visual grounding and applies corrections to the hidden states only when grounding deteriorates. The corrective update is computed analytically in closed form, requiring no training of auxiliary networks or model retraining. Experiments on LLaVA-1.5-7B and Qwen-VL-Chat show that BRACS consistently outperforms prior methods on hallucination benchmarks, reducing CHAIR

_s

by 9.4 points and improving POPE F1 by 2.7 points, while matching or improving performance on four general multimodal benchmarks. BRACS also remains efficient, operating at 80% of greedy decoding throughput and achieving 1.3 times higher speed on average than the baselines.

Soumyadeep Jana, Pulkit Mittal, Sanasam Ranbir Singh

Indian Institute of Technology Guwahati, India

Hallucination in Language Models Large Vision-Language Models Object Hallucination in VLMs Vision-Language Grounding Activation Steering VLM Hallucination Mitigation+1

May 28, 2026cs.CV

Multi-Stage VLM Pipeline for Zero-Shot Traffic Accident Understanding

We present the 1st-place solution to the ACCIDENT challenge at the CVPR 2026 AUTOPILOT Workshop, which asks for zero-shot prediction of accident timing, impact centroid, and collision type from CCTV footage. On a frozen Qwen3-VL-32B-Instruct checkpoint we build a three-stage pipeline (full-video joint prediction, time refinement, and single-frame grounding of the impact centroid), run the same pipeline a second time on a 235B Mixture-of-Experts sibling, blend the two outputs 9:1, and finally snap each predicted point onto the nearest vehicle detection. The final system reaches Public LB 0.55469 / Private LB 0.57080, roughly +0.21 over the strongest host baseline (Molmo-7B, 0.358) and wins the challenge. We ablate each component, report the negative results that shaped the final design, and release the code at https://github.com/fuumin621/cvpr2026-accident-1st-place-solution.

Fumiya Tatematsu, Fumihiko Takahashi

GO Drive Inc.

VLMs for Autonomous Driving Temporal Video Understanding VLM Inference