Mobile AI acts as a visual oracle, empowering users to snap a picture of something and ask for information. Snap-and-ask retrieval is now one of the most common entry points for mobile AI, yet photos are often blurry, while text questions may be short or mistyped. Existing benchmarks only test on clean inputs or do not isolate paired robustness in snap-and-ask retrieval. Therefore, we introduce SnapBench, the first paired benchmark for robust snap-and-ask multimodal retrieval, spanning 1,145 queries, 9,085 gallery items under 53 controlled corruption conditions with human annotations. We evaluate 16 multimodal retrievers, covering dual-tower encoders and embedding-based VLMs. Results show that image corruptions substantially degrade retrieval, while text corruptions mainly affect text-only retrieval and have limited impact on joint retrieval. Clean image-only retrieval often outperforms joint retrieval, indicating the coarse-text drag and the lack of cross-modal fallback under noisy inputs. SnapBench provides a controlled testbed for evaluating robust retrieval in snap-and-ask scenarios. We further propose MOOR (Modality-anchored, Outlier-aware, Optimal Reweighting), a simple adaptive fusion approach, highlighting the need for reliability-aware modality calibration in snap-and-ask retrieval.
Omni-modal large language models are increasingly evaluated on clean text--vision--audio inputs, where every channel is present, synchronized, and readily interpretable. Such scores are often taken as evidence of robust cross-modal fusion, but clean evaluation cannot tell whether success depends on stable cross-modal structure or on cues sufficient only in intact inputs. To address this gap, we define a modality fault line: a boundary at which model behavior becomes unstable when a modality remains present and human-interpretable, but its internal evidence structure is perturbed. We introduce SCEval (Structure-Corruption Evaluation) a diagnostic evaluation protocol that keeps the question, answer space, and modality channels fixed while applying controlled structural corruptions to text, vision, and audio individually and jointly. Built from $273$ human-verified tri-modal examples from Social-IQ, OmniBench, and VALOR, SCEval evaluates $15$ proprietary and open-source omni-modal systems. The results show that structural corruption lowers clean accuracy, text--vision damage forms the most stable shared fault line, and multi-modal degradation is non-additive rather than a simple function of the number of corrupted modalities. Clean omni-modal accuracy therefore does not establish that a model will remain reliable when cross-modal evidence becomes structurally unreliable.
Vision language models (VLMs) are increasingly used in industrial decision-making systems, such as recruitment support and recommendation. This motivates careful analysis of how VLMs process visual and textual information. In this work, we study how VLMs interpret text rendered as an image, and investigate the influence of visual styling biases. To this end, we introduce Stealth Visual Prompts, which subtly change visual styling of text, such as color and contrast, while preserving semantic content. Using these prompts, we systematically control the visual styling of words in text and measure their impact on the analysis performed by VLMs. We further analyze how such visual perturbations affect the latent representations of the vision encoder. From our experiments, we observed that coloring positive words in green consistently shifts sentiment predictions toward a positive direction. As a result, VLMs often fail to properly account for negative words present in the text. Our analysis suggests that this behavior is correlated with changes in the latent representations of the vision encoder induced by color variations. In addition, we show that reducing text--background contrast increases reliance on visually salient cues and leads to more incorrect Visual Question Answering (VQA) outputs. These results suggest that the visual styling of rendered text can guide VLMs' interpretation in ways that diverge from human semantic understanding. Project page: https://github.com/KohsukeIde/color-bias-vlm
Reinforcement Learning with Verifiable Rewards (RLVR) makes Multimodal Large Language Models more accurate, but the gains are brittle: simply paraphrasing a question or changing the prompt template can degrade them, which challenges reliable deployment in high-stakes scenarios like medical VQA. We trace this to two issues of the standard RL objective. First, the binary verifier conflates format with content, so the reward signal cannot tell a wrong answer apart from a misformatted one. Second, the training distribution covers only a thin slice of the real-world prompts that the model might meet at deployment, so policies that perform well on the training distribution can behave differently under unseen prompts during test. Both failures call for a robust post-training method that helps the policy cover a broader distribution of semantically equivalent prompts, and we identify two measures that help achieve this objective: separating format from semantics in the reward, and applying policy invariance across perturbed prompts with equivalent semantics. We therefore propose Prompt-Invariant RLVR (PIRL), consisting of a dynamic trinary reward and a consistency regularizer based on an embedding-space adversary. Under stress testing, PIRL's average accuracy on benchmarks drops by only $\le 1\%$, where GRPO drops ~3%. On dynamic evaluation, PIRL also achieves the smallest performance drop.
Multimodal intent recognition combines linguistic, acoustic, and visual evidence, but individual modalities may be noisy, missing, semantically conflicting, or disproportionately dominant. Existing methods typically infer modality importance implicitly and either reweight or suppress unreliable inputs, without determining whether a degraded modality can be repaired and subsequently trusted. We propose PRIME (Precision-weighted Reliability Inference and Modality rEstoration), a closed-loop reliability guided framework that jointly diagnoses, restores, and reassesses modality quality at the sample level. PRIME represents the weakness of each modality through a contextual log-variance estimated from complementary diagnostic evidence, including predictive confidence, epistemic disagreement, cross-modal consensus, and feature degeneracy. Because modality-reliability annotations are unavailable, the estimator is explicitly trained using controlled modality corruption with known degradation severity, together with a heteroscedastic uncertainty objective. Rather than directly discarding an unreliable modality, PRIME uses its estimated weakness to control a prototype-conditioned variational restoration module that reconstructs the degraded representation from complementary modalities. Crucially, reliability is re-estimated after restoration, allowing the model to determine whether the repaired representation has become sufficiently trustworthy to contribute to prediction. The resulting post-restoration precisions are used for inverse-variance multimodal fusion. Experiments on multimodal intent-recognition benchmarks show that PRIME maintains competitive clean-data performance while improving robustness under missing, noisy, conflicting, and modality-imbalanced conditions.
Multi-frame medical VQA appears to reward increasingly complex adaptation: controller-style inference, localization-aware reranking, static hard-negative mixing, and staged continuation all appear plausible from first principles. We test a simpler competing hypothesis on MedFrameQA: methods that remain tightly aligned with the benchmark's final answer objective should be the strongest \emph{robust} adaptation family once evaluation is controlled across fixed splits, matched budgets, repeated seeds, and calibration. We compare controller-based methods, scaffold evolution, static mixed supervision, continuation-heavy variants, and direct answer-only supervised fine-tuning (SFT). The strongest robust family is direct decoder-only answer SFT on MedGemma-1.5-4B. Empirically, this family yields substantial improvements in held-out report accuracy over frozen baselines while remaining remarkably stable across repeated seeds and matched controls, ensuring our claims reflect true family-level robustness rather than an isolated hyperparameter peak. Furthermore, post-hoc calibration effectively repairs confidence estimation without compromising accuracy, and the core approach transfers consistently to secondary backbones like Qwen2.5-VL-3B. The main result is therefore not that a complex auxiliary mechanism wins, but that objective-aligned direct answer SFT is the strongest robust adaptation family we found for MedFrameQA. By establishing this strong, minimalist baseline, we hope to redirect community focus toward fundamentally robust optimization rather than architectural complexity.
Francisco Mena, Dino Ienco, Roberto Interdonato +2cs.CV cs.AI cs.LG
Multi-modal classification leverages complementary information across diverse data sources to enhance predictive performance. However, real-world scenarios subject to operational constraints, such as sensor failures or privacy restrictions, lead to inconsistent modality availability between training and inference times. To handle missing modalities, prior studies have mainly covered bimodal data setups and focused on designing robust fusion processes. Instead, we adopt a multi-modal co-learning framework that prioritizes inter-modal collaboration rather than multi-modal fusion. Specifically, we consider that any subset of modalities may be absent, without assuming predefined missing-modality patterns, an inference scenario we refer to as missing arbitrary modalities. To address this challenge, we introduce two alternative approaches that leverage information at both feature- and decision-level. Experiments on two multi-modal classification benchmarks demonstrate significant robustness gains in various missing modality conditions. The first method shows more robust behavior under minimal missing conditions, where a single modality is absent, whereas the second performs better under extreme missing conditions, where all-but-one modalities are missing. Our code is available at https://github.com/fmenat/Co4Miss.
In this paper, we show for the first time that visual token pruning enhances the robustness of Multimodal Large Language Models (MLLMs), mitigating vulnerabilities such as jailbreak attacks and hallucinations. Given that vision and language modalities cannot be perfectly aligned, the misaligned visual tokens might act as out-of-distribution (OOD) inputs, leading to unpredictable outputs and introducing potential vulnerabilities. Building on this insight, we aim to enhance model robustness against jailbreaks and hallucinations by reducing OOD visual tokens at robust-pruning layers, while also reducing inference cost as a side benefit. Specifically, we measure the distance between each visual token and the language feature space. Then, visual tokens with large distances are identified as OOD tokens, which can be iteratively pruned. To demonstrate the effectiveness of our method, we evaluate it on seven diverse popular benchmarks. Notably, our method yields an average improvement of 13.29\% in defending jailbreak attacks, consistently achieves competitive performance in mitigating hallucinations, and maintains strong results on general datasets like MME.
Chengzhen Yu, Canran Xiao, Siyuan Ma +1cs.CV cs.AI
Vision-language alignment powers open-vocabulary recognition, retrieval, and LVLM grounding, yet natural captions are often underspecified, making similarity brittle and overly confident under paraphrase and omitted details. We aim to learn representations whose matching is stable across caption views and whose confidence reflects how strongly text constrains an image. We propose Text as Partial Constraint (TPC), a core-residual alignment framework that treats multi-view captions as incomplete supervision. It distills a consensus semantic core as the alignment target, learns a single-view core predictor for standard inference with one query, and explicitly discourages vision-language similarity from depending on the orthogonal unsaid residual. An uncertainty-aware contrastive objective further softens alignment when caption views disagree, reducing overconfident updates under weak language constraints. Across zero-shot recognition and adversarial robustness, TPC achieves 81.42/64.05 Top-1 clean/robust accuracy on ImageNet and 76.19/52.03 on an Avg-14 transfer suite, while improving LVLM transfer with 85.16 POPE F1 and 59.57 OKVQA accuracy under an LLaVA-1.5-7B stack. These results suggest that modeling text as a partial constraint is a practical and principled route to more reliable vision-language representations under underspecified language supervision.
Active Speaker Detection determines whether a visible person in a video is speaking at each moment. While recent audio-visual fusion methods perform well on clean data, they degrade under real-world corruptions such as background noise, occlusion, or simultaneous modality degradation. We attribute this limitation to the absence of explicit consistency constraints that promote robust, semantically aligned representations across modalities. Without such guidance, models tend to learn fragile modality-specific shortcuts that fail under corrupted conditions. We propose $C^3$ASD, a multi-level consistency-driven framework with three complementary constraints: embedding-level inter-modality consistency aligns audio-visual representations during speech; sequence-level intra-modality consistency separates speaking and non-speaking clusters via track-aware contrastive learning; and prediction-level consistency stabilizes fusion through knowledge distillation. Extensive experiments demonstrate significant improvements under diverse audio, visual and joint corruptions, while maintaining competitive performance on clean data.
Large Vision-Language Models (LVLMs) rely extensively on Visual Instruction Tuning (VIT) to elicit their multimodal reasoning capabilities. However, we find a discrepancy: VIT often packs multiple language tasks about the same image for conversational, multi-turn training, whereas existing benchmarks evaluate LVLMs in isolated, single-turn scenarios. The models can suffer from visual attention decay and contextual overfitting during multi-turn training, making it hard for them to realize their full potential in the mismatched test phase. To close the gap, we propose learning with Stochastic Turn Depth (StochasT), which stochastically groups language tasks for the same image into clusters of varying sizes (turn depth) while preserving their organic order. Hence, while StochasT draws on Dropout and stochastic depth for ResNets, it does not actually drop anything to maximize the utility of the training data. Furthermore, we introduce a challenging, benchmark-agnostic evaluation mechanism based on the Balanced Latin Square to measure LVLMs' robustness under varying contextual dependencies. Extensive experiments demonstrate that StochasT effectively grants LVLMs strong, harmonized capabilities for both single-turn and multi-turn use cases.
Video reasoning language models implicitly assume that every input frame is equally reliable. This leads to what we term the Blind Trust Problem: under realistic perturbations such as motion blur, glare, or occlusion, frontier video reasoning models can suffer 15-30%p accuracy drops on real-world embodied benchmarks, while remaining unaware that their visual evidence has been degraded. To address this challenge, we propose Robust-TO, an agentic video understanding framework that explicitly integrates per-frame trustworthiness into every stage of reasoning. Robust-TO organizes heterogeneous visual perception tools under a unified evidence interface. Each tool receives a sub-query derived from the original question and a set of trustworthy frames selected by the reliability-relevance score. It returns evidence in a shared format: a concrete prediction (e.g., a bounding box, motion trajectory, recognized text, or action label), temporal grounding, and a calibrated reliability score. During reasoning, these calibrated scores guide evidence weighting in a three-tier synthesis process (high/medium/low) and define a confidence-cost GRPO reward that jointly optimizes correctness, evidence reliability, and efficiency. On two video reasoning benchmarks spanning eight tasks, Robust-TO achieves 56.4% average accuracy on clean inputs, surpassing the strongest open-source baseline by 10.6%p and outperforming Gemini-2.5-Pro (46.2%). Under five realistic corruption types, Robust-TO maintains 54.3% average accuracy, 5.8%p above the strongest open-source baseline, while exhibiting the smallest clean-to-corrupted accuracy drop among all compared methods.
Ridwan Mahbub, Mohammed Saidul Islam, Md Tahmid Rahman Laskar +3cs.AI
Information visualizations are widely used to communicate patterns, trends, and outliers, yet deceptive design choices-such as truncated or inverted axes, distorted aspect ratios, inappropriate encodings, and misleading color mappings-can systematically alter interpretation while preserving the underlying data. As Vision-Language Models (VLMs) are increasingly used for chart understanding and analytical reasoning, assessing their robustness to such deceptive visualizations has become critical for trustworthy data analysis. We introduce VisDeception, the first controlled paired benchmark for evaluating the robustness of VLMs to misleading chart designs. The benchmark contains 1,600 paired faithful and misleading charts spanning eight major categories of deceptive visualization tactics, where each misleading chart is paired with a faithful counterpart generated from the same underlying data. To isolate deception-induced reasoning errors from baseline chart-understanding errors, we introduce the Deception Score, a paired evaluation metric that quantifies how misleading visualizations shift model responses away from the faithful interpretation of the data. Across 32,000 responses from 10 state-of-the-art VLMs, we find that even advanced models remain highly vulnerable to deceptive visual manipulations. To improve robustness, we further propose an inference-time multi-agent mitigation framework that grounds reasoning in structured chart metadata extracted from the visualization before answer generation, enabling models to reduce the influence of deceptive visual cues without requiring explicit user instructions. Together, our findings reveal important reliability gaps in current chart-understanding systems and establish benchmark-driven evaluation, deception-aware metrics, and structured reasoning as promising directions for developing more trustworthy VLMs for visual analytics.
Yinfeng Wang, Zhiyuan Yao, Zheren Fu +2cs.CL cs.CV
Multimodal large language models (MLLMs) are frequently exposed to auxiliary textual context, the impact of which on visually grounded tasks remains underexplored. In this paper, we investigate the influence of task-irrelevant context by formulating it as a controlled intervention within a binary visual judgment framework. By maintaining an invariant prompt structure while varying auxiliary inputs, we observe that irrelevant text consistently biases model predictions across diverse benchmarks. To move beyond performance metrics, we characterize this sensitivity through a decision margin defined by the log-probability difference between binary candidates. Our analysis reveals a robust geometric regularity: contextconditioned margins follow a consistent affine transformation of their context-free counterparts. This finding demonstrates that irrelevant context does not manifest as unstructured stochastic noise but as a estimable distortion of model preference. We further interpret the fitted affine parameters as metrics for visual commitment preservation and directional answer bias. These findings provide a margin-level diagnostic view of irrelevant-context effects in MLLMs and offer a basis for future studies on noisy-context robustness
Vision-language models (VLMs) are increasingly used for scene understanding in autonomous driving, but robustness analysis often relies on task-agnostic embedding stability alone. We study whether corruption-induced embedding drift predicts changes in a task-aligned hazard score derived from CLIP image-text similarities. Using controlled corruptions on BDD100K road scenes, we compare embedding drift against margin drift, defined as the change in hazard score under perturbation. The relationship is highly corruption-dependent: some families exhibit strong coupling between representation drift and decision drift, while others induce hazardous decision instability despite relatively modest embedding change. Furthermore, corruption families differ in failure direction: most suppress hazard detections via false negatives, while occlusion instead triggers false alarms, suggesting that benchmark design should account for asymmetric failure modes, not just overall instability rates. These results suggest that robustness benchmarks should include task-aligned stability measures in addition to embedding-level perturbation statistics.
Yizheng Sun, Mochuan Zhan, Yanan Ma +10cs.CV cs.CL
Reasoning Vision-Language Models (VLMs) achieve strong performance on complex multimodal tasks, but reliable real-world application requires handling visual inputs that are messier than clean, curated benchmarks. Existing works mainly evaluate such reliability of VLMs through input corruptions, such as noise, blur and weather effects, which make visual evidence harder to perceive. This leaves a critical reliability failure mode underexplored: a model may perceive the evidence correctly, yet reason from plausible but irrelevant and distracting evidence and propagate this mistake to its final answer. To address this gap, we introduce \textbf{Distract-Bench}, a benchmark for evaluating VLM robustness to \textbf{semantic visual distractions}, defined as meaningful but task-irrelevant visual cues added to inputs while preserving the ground-truth answer. We comprehensively evaluate eight leading open-source and two closed-source VLMs across conventional vision corruptions and Distract-Bench. Our results show that Distract-Bench exposes a robustness failure distinct from vision corruptions: reasoning VLMs largely track their non-reasoning base models under perceptual degradation, but show consistently lower robustness to semantic distractions. Further analysis shows that these distractions often enter the reasoning process of VLMs, are treated as evidence, and lead to incorrect answers. Together, these findings reframe robustness evaluation for reasoning VLMs, shifting the focus from degraded perception to distractions for reliable real-world visual reasoning. Our data and code are available at https://github.com/Yizheng-Sun/Distract-Bench.
Muhammad Falensi Azmi, Ikhlasul Akmal Hanif, Vallerie Alexandra Putra +3cs.CV cs.AI cs.CL
Symbolic benchmarks have emerged as a key approach to assess model robustness under minor modifications to STEM-related questions. However, existing symbolic benchmarks mostly remain limited to mathematical reasoning, lack visual grounding, and are predominantly in English. In this work, we introduce Sci-Rho (Science Rhobustness), a dynamic benchmark for visually-grounded STEM problems spanning five subjects and seven languages, comprising 4,242 problem templates (606 per language) crafted by domain experts, including Olympiad medalists. Each template is implemented as executable Python code that generates diverse but equivalent problem instances by varying numerical values, visual patterns, geometric shapes, color schemes, and function types, resulting in 42,420 instances in total, each paired with reasoning steps and ground-truth solutions. We evaluated 17 state-of-the-art VLMs and discovered a noticeable gap between worst-case accuracy (defined as the proportion of problem templates that a model answers correctly across every generated variation) and average accuracy. We also discovered that smaller models show noticeable performance degradation across languages, whereas proprietary and larger models remain robust. Step-level evaluation reflects this same trend, revealing a significant gap between average F1 and worst-case F1 scores. Finally, our inspection of attention heads of a VLM reveals substantial cross-lingual variation in the relative attention allocated to image tokens compared to text tokens. Our work highlights the importance of evaluation beyond static benchmarks as a metric to measure the quality of VLMs.
I Putu Adi Pratama, Bahadorreza Ofoghi, Atul Sajjanhar +1cs.CV cs.AI
Medical visual question answering (Med-VQA) has strong potential for clinical decision support by enabling AI models to interpret medical images and answer clinically relevant queries. Recent approaches typically connect off-the-shelf vision encoders with large language models (LLMs) through lightweight mapping networks to reduce computational cost. However, these methods often overlook the importance of handling noise and small irrelevant changes in visual representations. To address these challenges, we propose a noise-aware Med-VQA framework that incorporates a denoising autoencoder before visual embeddings are mapped into the input space of an LLM. The denoising autoencoder is pretrained to reconstruct clean visual embeddings from corrupted inputs, encouraging the model to learn robust visual representations that are less sensitive to noise. The resulting embeddings are then projected into the language model embedding space using a multi-layer perceptron (MLP), forming visual prefix tokens that provide image information to the LLM. To enable efficient adaptation without full retraining, we employ parameter-efficient fine-tuning using low-rank adaptation (LoRA). The proposed method is evaluated on the SLAKE and PathVQA benchmarks. Experimental results show improved robustness to noisy input embeddings while maintaining competitive clean performance across multiple evaluation criteria. These findings suggest that learning more robust visual representations can enhance Med-VQA performance and robustness.
Standard IMVC evaluation retrains separate models for different missing-data configurations. We show that this paradigm obscures a fundamental vulnerability: missing rate alone is insufficient to characterize data incompleteness. Specifically, we show that protocols with identical nominal missing rates can differ by up to $50\times$ in their proportion of fully observed samples, inducing drastically different learning regimes. We formalize this phenomenon as incompleteness divergence, providing measures that capture structural disparities across missing-data protocols. We further prove that for a broad class of reconstruction-based objectives, learning becomes structurally ill-posed when the proportion of complete samples falls below a critical threshold, leading to near-random performance. To bypass this theoretical bound, we propose CRAFT (Complete-data Robust Attention-masked Fusion Transformer). CRAFT shifts the burden of robustness from the loss function to the architecture via two key properties: (i) per-sample independence, which removes reliance on complete-sample co-occurrence, and (ii) mask-aware variable-length fusion, which aggregates only observed views through attention masking. This design allows a single model, trained once on complete data, to generalize to diverse missing patterns at inference time without retraining. Extensive experiments on seven benchmarks show that CRAFT matches or outperforms per-configuration baselines while reducing training overhead by $8.8\times$, demonstrating that robustness to missing data can be achieved as an inherent architectural property. Code (CRAFT) and our imvc-audit toolkit are available at https://anonymous.4open.science/r/CRAFT-BF80/ and https://anonymous.4open.science/r/imvc-audit-8263/.
Philip Wootaek Shin, Ajay Narayanan Sridhar, Sivani Devarapalli +3cs.CV cs.CL
Vision-language models (VLMs) achieve strong multimodal performance but remain prone to relation hallucination, which requires accurate reasoning over inter-object interactions. We study the impact of visual perturbations, specifically rotation and noise, and show that even mild distortions significantly degrade relational reasoning across models and datasets. We further evaluate prompt-based augmentation and preprocessing strategies (orientation correction and denoising), finding that while they offer partial improvements, they do not fully resolve hallucinations. Our results reveal a gap between perceptual robustness and relational understanding, highlighting the need for more robust, geometry-aware VLMs.