Vision-Language Models

Also known as VLM

Follow this topic Ask about this topic

Momentum

55 papers in the last four weeks, up 120% on the four weeks before. 0.5% of all new papers.

Jul 13Week of Sep 28

Latest papers 480

Oct 8, 2026cs.AI

Harness Compilation: Which Decisions Should a Small Vision-Language Model Keep?

Small vision-language models may be able to read external evidence yet struggle to obtain it. We introduce Harness Compilation (HC), an offline procedure that adapts the division of work between a frozen small VLM and its external harness. A large teacher uses student execution traces to revise reusable content and control, while a separate validation set selects the deployed harness. Deployment requires neither weight updates nor teacher calls. Across seven visual question-answering settings with students of at most 9B parameters, HC improves scores over bare students by 9.9-23.9 points, averaged over three independent builds per setting. Interventions on five runtime decision types (invocation, selection, argument generation, evidence integration and abstention) show why this allocation matters: requesting evidence and generating open queries can be costly, whereas bounded choices and reading supplied text can remain useful student work. Fact cards benefit all ten evaluated students, but decision policies transfer unevenly. Recompilation for a new student model helps when the transferred interface no longer fits the student. With 100 practice items, HC exceeds answer-only LoRA on three tasks. Larger training budgets can match or surpass a fixed harness, while combining the two improves SlideVQA beyond either alone. These findings support allocating work from measured student behavior rather than uniformly removing decisions.

Minhao Fan, Yinyi Liu, Jiayu Zhao +3

College of Computing and Data Science, Nanyang Technological University · School of Microelectronics, University of Science and Technology of China

Tool Use in VLMs Vision-Language Models Visual Question Answering

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 6, 2026cs.CV

SpaTime: Streaming Vision-Language Models for Spatio-temporal Reasoning

Embodied agents must reason about 3D space while the video is still arriving, answering questions as soon as they have observed enough of the scene. VLMs that incorporate 3D geometric priors achieve strong spatial reasoning, but they operate offline, i.e., the full video must be available before they produce an answer. Streaming VLMs process frames causally and decide for themselves when to respond, yet they lack explicit 3D representations. We present SpaTime, a streaming VLM that fuses causal geometry tokens into the language model at every frame, using only the frames observed so far. To supervise when the model answers, we propose a response-time loss that maps per-frame response probabilities to a differentiable expected response time and penalizes the distance from the ground-truth frame. For evaluation, we construct StreamVSTI-Bench and StreamVSI-Bench, streaming adaptations of VSTI-Bench and VSI-Bench. On StreamVSTI-Bench, SpaTime reaches 49.2% overall accuracy and reduces the mean response-time error by 66% relative to the strongest streaming baseline.

Hairong Yin, Huangying Zhan, Shin-Fang Chng +2

Department of Computer Science, Purdue University, USA · Goertek Alpha Labs, USA

Vision-Language Models Spatiotemporal Reasoning VLM Reasoning 3D VLMs Temporal Video Understanding

Oct 6, 2026cs.CL

Reading, Not Manipulating: Leveraging Router Logits for Multimodal Safety in MoE Vision-Language Models

Vision-language models (VLMs) face compositional safety risks where harmful intent emerges from the interaction between visual and textual inputs. As mixture-of-experts (MoE) VLMs become increasingly common, recent work has explored various safety interventions, including prompting, supervised fine-tuning, and routing-based expert steering. However, these methods show inconsistent improvements across models and evaluation distributions, and the intervention into model behavior or internal states introduce safety-utility tradeoffs by over-refusal. Rather than manipulating internal states to steer model behavior, we instead ask whether routing states can serve as diagnostic signals for multimodal safety. We find that router logits indeed provide highly predictive signals of whether a multimodal input is safe or not. Motivated by this observation, we introduce a lightweight router-logit safety detector that reads out routing signals during prompt prefill and identifies unsafe requests before generation, without modifying model parameters or expert routing. Across Qwen3-VL and Kimi-VL, the proposed detector substantially reduces safety errors on the HoliSafe benchmark and resoundingly generalizes to out-of-distribution safety benchmarks featuring different safety patterns, including MISHard and MM-SafetyBench. The success of the proposed router-logit detector also suggests a broader perspective on model internals: rather than focusing only on manipulating internal components to steer behavior, simply reading naturally emerging signals and linking them to an external safety mechanism can provide a simple, effective, and non-intrusive complement to existing safety interventions.

Ziyuan Yang, Wenxuan Ding, Shangbin Feng +1

University of Washington · New York University

Vision-Language Models Safety Filtering LLM Safety

Oct 5, 2026cs.CV

Semantic Capability Acquisition and Specialization During Vision-Language Model Fine-Tuning

Fine-tuning vision-language models (VLMs) is typically evaluated at a single downstream checkpoint, obscuring whether a semantic capability was never acquired or emerged earlier and later declined during specialization. We ask how semantic capabilities are acquired, when they peak, how well they transfer, and what remains at deployment. We study these dynamics as a semantic capability trajectory, tracking identity- and attribute-based capabilities over training. We formulate a trajectory-based framework that separates capability acquisition, capability-specific optima, and later specialization, and introduce Structured Semantic Routing (SSR) to study how the representation of supervision shapes what is acquired. Across six pretrained backbones spanning DFN, MetaCLIP, and OpenAI CLIP, we show that fine-tuning can acquire semantic capability beyond the pretrained state, including gains observed on held-out evaluations. Unstructured name-and-attribute supervision produces strong name-and-attribute retrieval with comparatively weak name-free attribute-profile retrieval, whereas SSR yields substantially stronger name-free attribute-profile retrieval and is further strengthened by stochastic name-branch dropout. Different capabilities can peak at different stages, so a checkpoint selected by target class-name retrieval need not coincide with a transferable semantic optimum. Continued optimization can therefore preserve strong target class-name retrieval while reducing previously acquired transferable semantic capability. In a representative diagnostic study, this late specialization is consistent with reduced cross-modal semantic accessibility while substantial image-only class structure remains available.

Suguru Onda, Matthew Bailey, Ryan Farrell

Brigham Young University · Carnegie Mellon University

Vision-Language Models VLM Adaptation Large Vision-Language Models

Oct 5, 2026cs.LG

Improving Proactive AI Assistance with Hierarchical Procedural Understanding

Proactive AI assistants continuously observe a user's activity and decide whether to provide new guidance or remain silent. They should provide appropriate guidance for the task, determine when to provide the next guidance based on task progress, and adjust the guidance level to the user's expertise and needs. Supporting these capabilities requires training and evaluation data that reflect procedural structure and capture how guidance should adapt to task progress and user needs. However, existing datasets either focus on detection-based proactive understanding or provide procedural guidance at a fixed granularity. Fixed-granularity guidance provides limited information about fine-grained progress and broader procedural context, making it difficult to determine completion and adapt guidance granularity. To address these limitations, we introduce the ProactiveCoach suite, comprising ProactiveCoach-Instruct for training, ProactiveCoachBench for evaluation, and fine-tuned VLMs with an adaptive guidance system. ProactiveCoach-Instruct provides hierarchically structured guidance at the phase, step, and action levels for learning task progress and procedural context. ProactiveCoachBench evaluates whether models provide appropriate guidance at the right time across different guidance levels and adapt when the requested level changes. We fine-tune pretrained VLMs on ProactiveCoach-Instruct and demonstrate its effectiveness across backbones. Compared with fixed-granularity supervision, hierarchical supervision improves overall performance across backbones by up to 9.6%p. We further build an adaptive guidance system by combining our fine-tuned model with a lightweight guidance router. Without additional fine-tuning, our system outperforms the in-context adaptation baseline by 57.1%p across four guidance-level transitions. Our project page is available at https://jinsuby.github.io/ProactiveCoach/.

Jin-Seop Lee, TaeYeon Won, SeongJun Jung +5

Department of Artificial Intelligence, Sungkyunkwan University, Republic of Korea · College of Computing and Data Science, Nanyang Technological University, Singapore

Vision-Language Models Proactive Assistance

Oct 5, 2026cs.CV

Fitting Vision Adapters at Frontier Scales

Training a small projector between a frozen vision encoder and language model is an established approach to multimodal learning. As the parameter count of language models scales dramatically, we revisit which vision capabilities this approach can add while keeping their pretrained weights fixed. Here we train a 50M parameter projector from the vision encoder of Kimi K2.6 to GLM 5.2 and 5.3, both models without native vision capabilities, and further present a reproducible recipe for training these adapters at scale. We study the following: (a) how vision capabilities of multimodal models scale as purely the language model side scales, and (b) what specific vision capabilities are able to be imbued into a pure language model at scale, and which ones remain limited. We evaluate on MMMU-Pro and BLINK, examining both overall performance and results on individual visual tasks.

Jaehoon Lee, Harry Partridge, Mudith Jayasekara +3

Base Labs

Vision-Language Models VLM Adaptation Large Vision-Language Models

Oct 4, 2026cs.CV

SPACE-CLIPv2: Decoding Local Geometry from Frozen CLIP for Monocular Depth Estimation

Vision-language foundation models such as CLIP provide strong semantic representations, but their patch tokens are not directly optimized for dense metric geometry. SPACE-CLIP showed that frozen CLIP features can support monocular depth estimation through layer-group feature fusion, yet it leaves open how neighboring CLIP tokens should be combined to recover fine local structure. We present SPACE-CLIPv2, a frozen-backbone depth decoder that aggregates fixed local neighborhoods in CLIP token space. At selected decoder stages, the model samples a fixed token stencil, predicts aggregation weights, and injects the resulting response through a gated residual update. A token-space high-pass branch further preserves shallow local contrast. On NYU Depth V2, SPACE-CLIPv2 improves over a matched SPACE-CLIP baseline, while five-seed experiments consistently favor fixed over learned-offset sampling. Zero-shot iBims-1 evaluation further improves boundary and planar-geometry measures. These results support constrained local token aggregation as a practical mechanism for decoding geometry from frozen CLIP representations.

Hyun Song, Taewan Cho, Kangmin Kim +1

School of Computing, iRASC Lab., Gachon University

Vision-Language Models Vision Foundation Models Monocular Depth Estimation

Oct 4, 2026cs.CV

LightVLN: Efficient Aerial Vision-and-Language Navigation with Compact Memory and History-Guided Local Aggregation

Aerial vision-and-language navigation (VLN) enables unmanned aerial vehicles to execute long-horizon natural-language instructions from visual observations in complex three-dimensional environments. However, recent aerial VLN models often rely on large-scale vision-language backbones and dense visual histories, imposing substantial computation and memory costs that hinder onboard deployment. We propose LightVLN, a lightweight history-aware aerial VLN framework that combines a compact 0.5B language backbone with compact representations of both historical and current observations. LightVLN compresses each historical frame into a single token using visual features already computed by the policy. It further introduces history- and instruction-conditioned local aggregation to reduce the current observation from 256 to 32 visual tokens while preserving navigation-relevant spatial information. With up to 16 historical frames, the policy uses at most 48 observation-derived tokens. On the public OpenFly dataset, LightVLN achieves 50.93% Test-Seen and 36.14% Test-Unseen success rates (SR), outperforming the evaluated 7B language-backbone baselines on most reported metrics. It also achieves 25.83% SR on AerialVLN-S Val-Seen. In a reconstructed unseen campus, we deploy LightVLN on a DJI M350 RTK with an external Jetson Orin NX 16 GB for closed-loop onboard-compute real-to-sim hardware-in-the-loop (HIL) evaluation, achieving 14.61 Hz model inference and 11.13 Hz end-to-end decision updates. These results demonstrate the effectiveness and efficiency of LightVLN for aerial navigation.

Yiming Zhao, Tianshun Li, Jingle He +2

The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China

Memory-Augmented VLMs Vision-Language Models UAV Navigation Vision-Language Navigation Aerial VLN

Oct 1, 2026cs.CV

Anti-Persona: Disrupting Unauthorized Identity Binding and Recognition in Personalized Vision--Language Models

Few-shot personalization enables large vision--language models (LVLMs) to learn user-specific visual concepts for applications such as personalized retrieval and subject-aware querying. However, it also creates a privacy risk: an adversary can bind a target identity from a few reference images and subsequently detect that identity in new images through natural-language queries. We introduce Anti-Persona, an image-level defense against unauthorized identity binding and recognition in personalized LVLMs. Our key insight is that identity personalization relies on visual features shared across multiple reference images. We aggregate these features into an identity prototype and optimize visually subtle perturbations that disrupt prototype alignment in the vision-encoder space. Spatial smoothing and low-frequency preservation further promote visual fidelity and practical resilience to image compression. The resulting protection does not depend on a specific prompt and supports both proactive anti-personalization and reactive image protection. Experiments on two representative personalized LVLMs demonstrate protection rates of up to

95.0\%

while preserving visual fidelity. The method remains stable across prompt variations and evaluated identity-query tasks, and improves black-box transfer under encoder mismatch.

Abhishek Basu, Fahad Shamshad, Karthik Nandakumar

MBZUAI, UAE · Michigan State University, USA

VLM Robustness Vision-Language Models Adversarial Examples Privacy Protection in VLMs

Oct 1, 2026cs.CV

Not All Error Yields to Scale: Where Scaling Stops in Vision-Language Inference

Vision-language models (VLMs) face a fixed-budget trade-off between processing more visual information for fine-grained perception and using a larger language backbone for complex reasoning. Existing studies do not tell us which combination of backbone size and input resolution to deploy, especially in high-resolution deployments. To address this gap, we propose the Separable Law that describes how VLM performance changes with language backbone size and visual token count. We fit the law to measurements from 26 InternVL and QwenVL models, with language backbone sizes from 1B to 72B, on four high-resolution benchmarks with image sizes from 224 pixels to 8K. We find that the questions responding to scaling can be predicted from the skill they require, while a substantial fraction never responds at all. We also find that the two model families gain similarly from a larger backbone, while their gains from more visual tokens differ sharply. Combined with a cost law, the Separable Law gives a closed-form rule for allocating compute between backbone size and visual tokens. When deployment is limited to available configurations, the law identifies model and image sizes that perform close to the best feasible choice under the same budget. We hope our work offers a principled way to decide how much a model should be allowed to see at high resolution, given what it must reason about.

Xinye Zhao, Yunkai Dang, Yunchen Wu +1

School of Intelligence Science and Technology, Nanjing University

Vision-Language Models Efficient VLM Inference Language Model Scaling Laws

Oct 1, 2026cs.CV

Fusing Visual and Textual Representations via Multi-layer Fusing Transformers for Vietnamese Visual Question Answering

In recent decades, artificial intelligence has made significant progress in understanding and interacting with images. One of the important applications of this technology is Visual Question Answering (VQA), a research field that requires computers to understand and answer questions about images in a natural manner. Despite extensive research and development in VQA for English, there have been very few similar efforts made for other languages, especially Vietnamese. This gap presents a significant challenge and opportunity for the advancement of VQA technology in the Vietnamese language context. By bridging this gap, the field of Vietnamese VQA not only enriches the diversity of research in artificial intelligence but also enables practical applications in various domains, such as education, healthcare, and entertainment, catering to Vietnamese-speaking populations worldwide. Thus, the exploration and development of Vietnamese VQA systems hold immense potential for advancing both research and practical applications in the intersection of computer vision and natural language processing. In this paper, we propose a Multi-layer Fusing Transformer model utilizing a cross attention module to combine multiple modality features of images and texts from different layers in an aggregated representation. Our architecture allows us extract information from low level to high level. Through detailed experiments and ablation studies, our model achieves promising results against the competitive baselines in ViVQA dataset for Vietnamese language.

Cong Phu Nguyen, Huy Tien Nguyen, Tung Le

Faculty of Information Technology, University of Science, Ho Chi Minh City, Vietnam · Vietnam National University, Ho Chi Minh City, Vietnam

Visual Question Answering Vision-Language Models Cross-Modal Attention

Oct 1, 2026cs.AI

AiSearch: Interactive Multi-Modal Search with VLMs

Modern retrieval systems must both be automated and interactive, allowing users to search and refine results in real time. We present AiSearch, a flexible multimodal retrieval framework that leverages the zero shot capabilities of Vision Language Models (VLMs) for natural language search over images and videos. AiSearch supports interactive search refinement through user feedback to tailor results to the user's intent, and allows visual benchmarking across multiple VLMs, enabling users to select the most suitable model for their task.

Ali Koksal, Mei Chee Leong, Vicky Sintunata +2

Institute of Advanced Intelligence and Computing (IAIC), Agency for Science, Technology and Research (A*STAR), Singapore

Multimodal IR Vision-Language Models Information Retrieval Vision-Language Retrieval

Oct 1, 2026cs.CV

FlashBack: Knowing When to Remember in Streaming Vision-Language Models

Streaming vision-language models must process continuously growing video streams under a bounded compute budget, creating a persistent tension between real-time perception and long-term memory. Retrieving historical information provides a natural remedy, yet historical recall is not uniformly beneficial: unnecessary history may introduce irrelevant context into current reasoning and interfere with native real-time perception. Effective streaming memory should therefore address not only what to remember, but also when and how to access it. To this end, we introduce FlashBack, a training-free framework for selective, multi-level memory in streaming vision-language models. Before retrieving history, FlashBack draws on the semantic understanding of the frozen streaming VLM to infer whether a query calls for historical evidence. This assessment determines whether inference remains on the Native trajectory or invokes an isolated Recall trajectory. The Recall trajectory combines recent context with retrieved long-term memory through a query-local Side-KV pathway, preserving local temporal continuity without modifying the persistent Native state. We instantiate FlashBack on StreamingVLM and Mage-VL-4B and evaluate it on OVO-Bench and StreamingBench. The results show improvements on several long-horizon and memory-dependent tasks while largely preserving real-time perception, with performance competitive with strong training-based streaming methods despite requiring no additional training. Our code will be announced later.

Yi Chen, MingMing Yu, Rui-Qi Wang +5

Rightly Robotics, A4X

Memory-Augmented VLMs Vision-Language Models Persistent Memory for Language Models Temporal Video Understanding

Oct 1, 2026cs.CV

Geometric Similarity in VLM Low-Level Vision Representations

Vision-language models (VLMs) have emerged as powerful candidates for universal vision backbones, with representative architectures including autoregressive (AR) models and diffusion transformers (DiTs). Yet, adapting them efficiently for all-in-one low-level image restoration remains a challenge. Crucially, the field lacks an understanding of how VLMs organize hidden-layer representations and whether these structurally distinct paradigms share a common geometric organization for pixel-level perception. Such shared organization is a prerequisite for building highly transferable, unified restoration VLMs and adapters. In this paper, we systematically investigate representational similarity across 24 low-level tasks spanning 5 categories. We propose GeoSim, a unified four-level framework that analyzes task-conditioned representations from global similarity, local geometry, sparse feature decomposition, and topological verification perspectives. Our formulation applies to the analysis of hidden states in AR models and feature maps in DiTs across same- and cross-task/model settings. Our results reveal the organizing principles of low-level visual representations while exposing their limits in cross-task and cross-model agreement. Ultimately, GeoSim provides an interpretability lens for probing latent transferability in low-level vision and diagnosing model limitations in task- or model-specific scenarios.

Shao-Jun Xia, Huixin Zhang, Zhen Lei +3

Florida State University · UNC Chapel Hill

Vision-Language Models Representational Similarity Analysis Neural Representation Geometry Image Restoration

Sep 30, 2026cs.CV

Is Better Teacher Supervision Enough? Unlocking Student-side Learning in Multimodal On-Policy Distillation

On-policy distillation (OPD) improves reasoning by providing token-level supervision from a teacher on a student's own trajectories. Existing methods primarily focus on enhancing this teacher-side guidance (e.g., by enriching teacher inputs and refining teacher feedback), yet we find that limited student perception is another critical bottleneck in multimodal OPD. By providing oracle visual facts, the performance of OPD-trained students can still be substantially improved for both weak and strong teachers. To address this bottleneck, we propose S-OPD, a simple multimodal on-policy distillation framework that explicitly strengthens student perceptual learning through two objectives. Specifically, Teacher-calibrated Policy Contrast separates student policies under original and masked images with teacher-based token-level gating, strengthening the student's reliance on visual evidence during reasoning. Policy Agreement aligns student policies under original and noise-perturbed images, further improving perceptual robustness to visual noise. Notably, our method can be seamlessly plugged into existing OPD frameworks, requiring no additional data annotations, model parameters or inference operations. Extensive experiments on eight benchmarks across student scales and distillation paradigms demonstrate consistent performance improvements, with gains of up to 4.25 points on LogicVista. When combined with existing teacher-side supervision methods, our method can yield further gains. Code is available at https://github.com/Sirilaw/S-OPD.

Siyuan Liu, Kanghui Tian, Yue Duan +4

Nanjing University · Fudan University · Nanjing University of Posts and Telecommunications

Vision-Language Models Visual Reasoning On-Policy Distillation

Sep 30, 2026cs.CV

CRAFT: Causal Responsibility and Failure Tracing in Medical Vision Language Models

As vision language models are increasingly deployed in clinical diagnosis, understanding how they internally resolve competing visual and textual signals becomes a safety imperative. Existing mechanistic analyses remain confined to unimodal text and offer no explanation for why a single misleading sentence can override a correct image based diagnosis, or why a model commits to a confident answer despite insufficient visual evidence. We find that these two safety risks, arbitration failure where textual context overrides visual grounding and brake failure where the model commits without adequate evidence, are mediated by spatially disjoint attention head populations: arbitration heads form a mid-to-deep wideband reflecting cross-layer evidence competition, while brake heads concentrate in a narrow middle-to-late layer band that regulates evidence sufficiency and abstention behavior. To ground these observations in causal circuitry, we introduce CRAFT, which localizes each failure mode to a minimal causal head set via dual criteria and verifies necessity and sufficiency through temporal probes and Tuned Lens trajectory analysis. Excising arbitration heads sharply reduces conflict following with negligible degradation on clean inputs, while excising brake heads restores appropriate abstention under degraded visual evidence. The two interventions target spatially disjoint head sets and produce distinct corrective effects, underscoring the mechanistic separability of the failure modes. Experiments across multiple medical VQA benchmarks and VLM architectures validate both the localization and interventions, demonstrating that the identified heads causally drive each failure mode and that targeted modulation generalises without retraining. The code is available at GitHub repository.

Chunzheng Zhu, Jiaqi Zeng, Hongbo Zhao +3

Hunan University · University of Hong Kong

Vision-Language Models Medical VQA VLM Interpretability Medical VLMs

Sep 29, 2026cs.CV

LoopVL: Recurrent Visual Intelligence

We introduce LoopVL to study whether Loop Transformers can be effectively extended to vision- language models. LoopVL combines Module-Loop and Model-Loop computation to iteratively update a unified vision-language state through shared modules. We train LoopVL from scratch through language pre-training, multimodal training, and post-training. LoopVL outperforms a range of similarly sized and larger non-recurrent models on multimodal understanding and visual reasoning benchmarks. We also observe Visual Aha Moments in LoopVL, characterized by pronounced shifts in visual attention across loops. LoopVL provides practical evidence for recurrent vision-language modeling and offers an intuitive perspective on how shared parameters can support deeper multimodal computation over continuously evolving visual-language states.

Zhe Qian, Ziyang Gong, Zhongxing Xu +9

Gaoling School of Artificial Intelligence, Renmin University of China · Baidu · Shanghai Jiao Tong University +5

Vision-Language Models Large Vision-Language Models Recurrent Transformers VLM Reasoning

Sep 29, 2026cs.CV

Inductive Visual Logic for Few-Shot Out-Of-Distribution Adaptation in VLMs

Generative vision-language models (VLMs) such as Qwen-VL and LLaVA achieve strong zero-shot performance on tasks overlapping with their pretraining distribution, yet fail on specialized domains where the required discriminative features were never learned, a regime we term distant out-of-distribution (OOD). Standard adaptation methods cannot overcome this representational absence because they operate within the encoder's existing feature space. However, VLMs retain a robust descriptive capacity even when discrimination collapses: a model that cannot classify a medical scan can still articulate its visual patterns. Exploiting this asymmetry, we introduce Inductive Visual Logic (IVL), a training-free framework that constructs classification knowledge from the model's surviving descriptive ability. IVL extracts visual traits from few-shot support images through dual-mode prompting, combining semantic descriptions with primitive visual observations, and organizes them into per-class trait dictionaries. At inference, hierarchical filtering identifies spatially grounded trait evidence for classification. Across multiple distant-OOD benchmarks, IVL achieves the highest aggregate accuracy under two VLM backbones while producing interpretable, trait-traceable predictions.

Hung-Jen Chen, Yu-Heng Ho, Ting-Yao Huang +4

National Tsing Hua University, Hsinchu, Taiwan · National Taiwan University, Taipei, Taiwan

Vision-Language Models Few-Shot Domain Adaptation VLM Adaptation OOD Generalization

Sep 29, 2026cs.CV

It Takes Little to Rewrite Perception: Targeted Semantic Substitution in Vision-Language Models at $ε\leq 4/255$

Vision Language Models (VLMs) are widely deployed in safety-critical scenarios, and understanding to which extent they can be controlled by adversarial perturbation is a prerequisite for evaluating their trustworthiness. Existing representation-alignment attacks, which make a VLM perceive a target image, achieve limited success at

\varepsilon \leq 4/255

. Therefore, VLMs seems robust to perturbations in this range. We show that this robustness does not hold, as targeted semantic substitution succeeds within the same range. Specifically, we align each stream of the source image with its counterpart in the target image in the victim VLM's post-merger token space, operating under a white-box threat model. We evaluate under a strict success criterion, requiring the model to simultaneously name the target, confirm its presence, and deny the source. In images, target semantics appear at

\varepsilon = 2/255

and complete replacement reaches 38% at

\varepsilon = 4/255

. On video, complete replacement reaches 35.9% at

\varepsilon = 1/255

. We also observe a phenomenon of \textit{semantic fusion}, where Large Language Model (LLM) rationalizes contradictory visual signals into a coherent narrative.

Binchi Zhang, Atrisha Sarkar, Apurva Narayan

University of Western Ontario

VLM Robustness Adversarial Attacks on VLMs Vision-Language Models

Sep 29, 2026cs.CL

From Routing Signals to Selective Review: Visual regrounding in MoE VLMs

Vision-language models (VLMs) may accept false visual premises, answering questions about a target object's color, count, location, or state even when it is absent. We call this reliability-critical behavior a target-absence grounding failure. Existing visual-grounding detectors primarily rely on generated responses, hidden states, or uncertainty measures. We present the first framework to leverage internal routing decisions in Mixture-of-Experts (MoE) VLMs to detect target absence before generation and guide selective correction. We extract target-token routing probabilities from Qwen3-VL-30B-A3B-Instruct and Gemma-4-26B-A4B-it, train a separate L2-regularized linear detector for each model, and use its predictions to selectively invoke a target-aware review prompt. Using routing alone, the Qwen and Gemma detectors achieve ROC-AUCs of 0.9988 and 0.9956 on GQA-Inpaint and retain 0.8095 and 0.7781 on the external OBER dataset, respectively. The resulting routing-gated policy improves end-to-end accuracy on GQA-Inpaint and OBER by +22.25% and +12.17% for Qwen, and by +13.42% and +1.39% for Gemma, without modifying model weights. Further analysis shows that the signal is localized to the target-object token, emerges in early MoE layers, and is distributed across partially substitutable experts. Although cross-dataset threshold shifts require recalibration, false-positive review causes limited harm overall, suggesting that intervention risk can be controlled through joint selection of the detector threshold and review prompt. Overall, we show that routing probabilities alone preserve actionable information about visual perception, allowing computation already produced by an MoE VLM to support low-cost detection and selective visual regrounding.

Hongzhu Guo, Mohsen Fayyaz, Nanyun Peng

University of California, Los Angeles · Peking University

Hallucination Detection in VLMs Vision-Language Models Object Hallucination in VLMs

Sep 29, 2026cs.CL

Can Vision-Language Models Stay Helpful When Facing Implicit Risks? Intent-Privilege OPSD for Efficient Safety-Helpfulness Alignment

Vision-Language Models (VLMs) remain vulnerable to cross-modal implicit risks: visual and textual inputs that appear benign in isolation can jointly elicit unsafe responses. Existing safety methods often require large preference datasets, costly multi-rollout training, or additional safeguards at inference time. They may also sacrifice helpfulness by directly refusing requests that could be answered safely. In this paper, we propose Intent-Privilege On-Policy Self-Distillation (OPSD), which leverages evidence-grounded intent as privileged supervision during training to help VLMs recognize implicit risks and provide safe, useful responses instead of blanket refusals. OPSD distills a teacher's intent-conditioned preferences over responses into a student using a single rollout per prompt; the student then responds without intent annotations or an additional safety module. With only 1,447 safety-specific examples - 95% fewer than standard preference datasets - OPSD reduces training time by 5x relative to multi-rollout GRPO-style training and average inference length by 7%. It attains the highest ratio for joint safety-helpfulness success, which measures the proportion of responses that are both safe and helpful, across all five evaluation groups. Remarkably, on pooled SIUO+HoliSafe, this success ratio rises from 43.9% to 53.5%. These results show that training-time intent supervision can improve both safety and helpfulness while substantially reducing data, training, and inference costs.

Haotian Deng, Wenbin Xing, Gang Xu +5

Southern University of Science and Technology · Sun Yat-sen University · Guangdong Laboratory of Artificial Intelligence and Digital Economy (SZ) +5

VLM Robustness Vision-Language Models On-Policy Self-Distillation

Sep 29, 2026cs.CV

Are In-Context Images Worth 10 Dimensions?

There has been significant work on understanding the In-Context Learning capabilities of Large Language Models, especially on the induction circuit. For a few-shot classification task, the induction circuit leverages linear representations of each labeled example in-context in order to classify an unlabeled query. However, few works focus on how those linear representations are built in the first place. Leveraging the expressivity of the vision modality compared to text, we uncover a Shared Discriminative Geometry (SDG) inside Large Vision Language Models (LVLMs). It is a low-dimensional space, shared across all image classification tasks, in which in-context images are compressed into linearly separable representations later used to perform classification. We observe that this is the result of the model performing a dimensionality reduction of vision representations in early layers. In order to explain this phenomenon: (1) We show analytically that linear self-attention can perform a dimensionality reduction by projecting in-context data onto its principal components, with each layer implementing one gradient descent step toward this objective. (2) We provide evidence that trained LVLMs reduce the dimensionality of vision representations in early layers via a similar mechanism.

Adhemar de Senneville, Xavier Bou, Jérémy Anger +2

Université Paris-Saclay, CNRS, ENS Paris-Saclay, Centre Borelli, Paris, France · Pôle recherche de l’AMIAD, Palaiseau, France · Institut Universitaire de France, Paris, France

Vision-Language Models Multimodal ICL Large Vision-Language Models VLM Interpretability

Sep 29, 2026cs.CV

Targeted Visual Counterfactual Explanations for Contrastive Vision-Language Model

Current explanation methods for contrastive vision-language models such as CLIP mainly identify important regions without showing how to change the input in order to get a target prediction. We introduce Mask-guided Adaptive Counterfactual Explanations (MACE), a targeted visual counterfactual method designed specifically for CLIP zero-shot classification. MACE constructs an editable region from either source attribution or source-target attribution differences and expands the mask only when needed to reach a specified target class. A latent diffusion inpainting model then modifies the selected region, while a frozen CLIP model provides modification guidance and anchors the remaining image content to the original input. We evaluate MACE on ImageNet, Food-101, Oxford Pets, and CUB-200. The source-mask variant achieves the highest target top-1 success rate across all four datasets, while the difference-mask variant produces the smallest pixel level and perceptual changes and the best realism scores. Both variants improve proximity and realism over a Stable Diffusion-only baseline using the same generative backbone. These results show that adaptive mask-guided editing produces effective CLIP counterfactuals. They further reveal a tradeoff between counterfactual validity and source-image preservation.

Van Bach Nguyen, Jörg Schlötterer, Christin Seifer

Marburg University, Germany

Vision-Language Models VLM Interpretability

Sep 29, 2026cs.CV

Beyond Attention Imbalance: Mitigating Hallucinations via Spectral Surgery

While Large Vision-Language Models (LVLMs) achieve remarkable success, hallucinations remain a significant barrier to their reliable deployment. Recent studies primarily attribute these issues to cross-modal attention imbalances; most solutions therefore focus on reweighting visual tokens or suppressing language priors. However, such approaches often overlook the spectral characteristics of the visual information flow and frequently rely on Contrastive Decoding (CD), which doubles inference time. Instead of following conventional approaches, we identify two distinct hallucination patterns-Perceptual-Semantic Dissociation and Localized Fixation-and propose FLASH (Frequency-Localized Attention SHaping), a training-free and CD-free framework. FLASH utilizes a Spectral Vortex Score to detect vision heads within multi-head attention layers and applies adaptive spectral modulation to rectify the visual information flow during decoding. Empirical results demonstrate that FLASH achieves a superior balance between performance and efficiency compared to SOTA methods.

Siqi Lu, Suo Wei, Yongbin Zheng +3

College of Intelligence Science and Technology, National University of Defense Technology, China · School of Computer Science, Northwestern Polytechnical University, China · Ningbo Institute, Northwestern Polytechnical University, China +1

Vision-Language Models Efficient VLM Inference Large Vision-Language Models VLM Hallucination Mitigation

Sep 29, 2026cs.CV

Seeing Is Not Addressing: Auditing Linguistic Access to Frozen Visual Geometry

Visual distinctions are often finer than those reflected in linguistic conceptualization. Vision-language models exhibit a similar asymmetry: a distinction can remain discriminable in frozen image geometry while being weakly addressable through the native text interface. We study this gap by separating visual discriminability from linguistic addressability in text-to-image retrieval. Using FactorAtlas, a fully crossed testbed of 23,040 images spanning shape, hue, pattern, and nuisance variation, we compare both readouts on held-out images of the same distinctions. We then derive image-side contrasts that separate each value from its alternatives for matched visual grounding, and test whether this reduces the native-text access gap across factors and models. Direction-specific and visual-absence controls tie these gains to the relevant visual contrast; the gains persist after global alignment and extend to compositional retrieval and natural images. Together, these results show that visual discriminability and linguistic addressability need not coincide, and that matched visual grounding can probe and reduce the resulting access gap.

Woosang Jeon, Jiwon Yang, Soo Chung +1

Seoul National University

Vision-Language Models Vision-Language Grounding Modality Gap in VLMs Vision-Language Retrieval Fine-Grained Visual Perception

Sep 29, 2026cs.CV

Exploring In-Context Learning for Handwritten Text Recognition

Handwritten Text Recognition (HTR) systems have become an indispensable tool for the digitization of historical documents. Not only do they cut down time and cost, but they also allow democratizing access and processing of their contents by generating their transcripts. However, literature in HTR currently focuses mostly on specialized models that require large amounts of annotated samples to achieve satisfactory performance. We explore the use of In-Context Learning with pre-trained Vision-Language Models (VLMs) to create a transcription pipeline without updating the model's parameters. We then evaluate this pipeline across multiple collections and models, and demonstrate that general-purpose VLMs can be effectively taught how to transcribe handwritten text from images. To assess how our observations may translate to practical applications, we evaluate the performance in a Cross-Domain (CD) scenario, where context examples are drawn from a different collection than the query image. Results in both the controlled In-Domain (ID) scenario and the realistic CD scenario follow the same patterns. First, as context size grows, the error range is expected to narrow towards the average performance. Thus, larger context sizes sacrifice the performance of the oracle-best sampling for lower expected error rates. The results obtained show that, without any parameter updates, this methodology has strong potential to compete with traditional HTR in the presence of domain shift. Moreover, we show and argue that some context samplings work better than others and suggest more effort should be put into finding an ideal sampling method in future work.

Eric Ayllon, Abel Gandia, Jorge Calvo-Zaragoza

In-Context Example Selection Vision-Language Models Multimodal ICL Cross-Domain Generalization

Sep 29, 2026cs.CV

DARE to Mitigate Hallucination: Dual-path Auto-Regressive-aware Editing

Large vision-language models (LVLMs) have recently achieved remarkable progress across multimodal tasks, yet object hallucination remains a persistent challenge where models generate descriptions inconsistent with the visual input. Recent work mitigates hallucinations through training-free representation editing, typically by constructing hallucination-related directions from teacher-forcing (TF) contrasts between hallucinated and truthful responses. However, LVLMs operate through autoregressive (AR) decoding during generation, raising the question of whether TF-based analysis fully reflects the generation dynamics that lead to hallucinated outputs. In this paper, we analyze the relationship between TF-based editing and AR generation behavior and find that TF-based editing alone may be insufficient to capture both decoding dynamics and multimodal interactions associated with hallucinations. To address this limitation, we propose DARE (Dual-path Auto-Regressive-aware Editing), a hybrid hallucination editing framework that integrates two complementary contrast pathways: textual contrasts and image contrasts, together with autoregressive-aware representation signals. Specifically, DARE constructs hallucination editing directions from (1) TF-based textual contrasts, (2) AR-aware representation transitions during decoding, and (3) controlled visual differences between paired images. Extensive experiments on multiple LVLM hallucination benchmarks demonstrate that DARE consistently reduces object hallucinations while preserving multimodal perception capability and inference efficiency. Our implementation code is available at https://github.com/KU-VGI/DARE.

Jae-Ho Lee, Jeong-Eun Lee, Gyeong-Moon Park

Korea University, Seoul, Republic of Korea

Vision-Language Models Large Vision-Language Models Object Hallucination in VLMs VLM Hallucination Mitigation

Sep 28, 2026cs.CV

Xiaomi-OCR-0 Technical Report

Compact OCR-specific vision-language models achieve strong document parsing performance, but often rely on costly supervision and focus primarily on visual-text reconstruction. We introduce Xiaomi-OCR-0, a unified 0.8B model for document parsing and OCR-centric understanding. We build an approximately 170M-sample OCR-centric corpus using an automated data engine that combines expert consensus, render-based verification, and targeted synthesis. Starting from Qwen3.5-0.8B, our progressive training recipe combines Q-Mask-based text anchoring, continued pretraining, and mixed-task reinforcement learning (Mix-RL). Xiaomi-OCR-0 achieves 95.24 on Real5-OmniDocBench, 96.83 on OmniDocBench v1.6, and 87.94 on Wild-OmniDocBench, while reaching an average score of 83.2 across five OCR-oriented VQA benchmarks. Ablations further show that, with sufficient parsing training, OCR-centric understanding supervision provides additional gains for document parsing. Homepage: https://huggingface.co/spaces/SeerRay-Lab/Xiaomi-OCR-0.

Xin Chen, Anan Du, Feng Feng +7

SeerRay Team

Vision-Language Models Document Parsing Optical Character Recognition Document Image Understanding

Sep 28, 2026cs.LG

Narrow Multimodal Fine-Tuning Can Induce Emergent Misalignment

Modern AI models are aligned through post-training to adapt them to downstream tasks. Recent work shows that fine-tuning language models on narrow tasks can induce emergent misalignment (EM), causing broadly harmful behaviors beyond the training task. However, EM has been studied almost entirely in text-only tasks, leaving its manifestation in multimodal models unclear. In this paper, we define and analyze EM in the context of vision-language models. We first induce EM via fine-tuning on narrow multimodal tasks targeting vulnerable code, careless household-object use, and conspiratorial interpretations of ordinary scenes. Across fifteen commercial and open-source models with different scales, we find that narrow multimodal fine-tuning can induce coherent and broadly misaligned behavior that transfers to unrelated tasks, including misaligned opinions, visual factual dishonesty, unsafe image generation, vulnerability to visual jailbreaks, and risky agentic actions. We further find that multimodal EM does not depend on the apparent harmfulness of training data but is sensitive to training-evaluation modality alignment. EM can arise under both supervised fine-tuning and preference optimization and can propagate through intermediate reasoning. Finally, we explore several mitigation strategies, including prompt inoculation, benign continued training, and activation-level steering, which can partially reduce EM. Overall, our findings suggest that multimodal EM reflects a behavioral shift rather than a general loss of capability, extending beyond text to the visual modality.

Shunchang Liu, Lukas Fluri, Xin Chen +1

ETH Zürich · EPFL · Aalto University +1

Vision-Language Models Fine-Tuning Emergent Misalignment in Language Models