VLM Evaluation

VLM: Vision-Language Model

Follow this topic Ask about this topic

Momentum

70 papers in the last four weeks, up 67% on the four weeks before. 0.7% of all new papers.

Jul 13Week of Sep 28

Latest papers 649

Sep 29, 2026cs.CV

Composition, Not Conversation: VLMs Lose the Scene, Not the Thread

Vision-language models (VLMs) increasingly reason over visual evidence that is cropped, segmented, retrieved, or revealed over time. Yet most VQA benchmarks present the complete image and question at once. We ask what models lose when the same information is fragmented. We introduce Layered-VQA, with 93 scenes and 300 questions. Each image is decomposed into ordered RGBA layers that exactly recompose the original scene, and each question is annotated with supporting, minimal-sufficient, and distractor layers. We evaluate eleven open-weight VLMs from 3B to 32B parameters and two proprietary models with a scale of 187,200 conversations, graded by 1.74M open-model cross-judgments. We find three consistent failures. Loss in Composition: fragmenting the question has a small effect, but fragmenting the scene substantially reduces accuracy; recomposing the same layers largely restores performance. Oracle Inversion: even oracle-selected sufficient evidence can perform worse than the complete scene. Loss in Grounding: as more evidence is required, grounding degrades much faster than answer accuracy. Together, these results show that having the right visual evidence is not enough. How that evidence is composed and presented determines whether models can use and ground it. The right evidence is not enough: VLMs need the scene it came from.

L. D. M. S. Sai Teja, Ufaq Khan, N. Siva Gopala Krishna +5

NIT Silchar · MBZUAI · BML Munjal University +1

VLM Evaluation Visual Question Answering Vision-Language Grounding VLM Reasoning

Sep 29, 2026cs.CL

It's Not What the Image Shows: Irrelevant Context Destabilises VLM Judges Without Informing Them

Vision-language models (VLMs) are increasingly used in place of human annotators, making it important that substitutability tests reflect the model rather than incidental evaluation conditions. We introduce MIST, the Misleading-Image Stress Test: 200 English sentences, each built around a phrase readable either figuratively or literally and shown with an aligned image depicting its reading, a misleading image depicting the opposite, or no image at all. The guidelines require the label to be decided from the sentence alone, so no image should change any answer. We expected each image to pull a judge's labels toward the sense it depicts, and neither kind did. Across thirteen VLM judges, an aligned image changed 20.5% of labels and a misleading one 19.4%, close for every judge and both above the 11.6% produced by deleting the ignore-the-image instruction with the image left in place. Yet only 37% of the labels that differ between the two images moved toward the sense shown, and agreement with our human annotators is unchanged whether the image is absent, aligned or misleading. The effect is smaller in the seven judges that pass the alt-test than in the six that never do, but present in all of them: what moves a judge is that an image is there, not which of the two it is, so a substitutability verdict describes a configuration as much as a model.

Nagham Omar, Mahmoud Jabarin, Kinan Ibraheem +1

Faculty of Data and Decision Sciences, Technion

VLM Evaluation Figurative Language Understanding VLM Robustness Counterfactual Evaluation of VLMs Vision-Language Model-as-a-Judge

Sep 29, 2026cs.RO

Faster and Better? Benchmark Bugs and Design Limitations Distort the Evaluation of Vision-Language-Action Acceleration

Simulated manipulation benchmarks are the standard tool for evaluating vision-language-action (VLA) policies and the acceleration methods that reduce their inference latency for on-robot deployment. On these benchmarks, we observe that some training-free acceleration methods, which approximate the baseline policy's computation, achieve higher measured success rates than the baseline itself. Success rates alone cannot establish whether such gains come from better task execution or from evaluation flaws. We therefore investigate two kinds of benchmark flaws behind these gains: bugs, where the implementation does not match the intended task or evaluation protocol, and design limitations, where success criteria and simulation settings do not fully capture how acceleration affects task execution. Starting from tasks with anomalous gains, we localize root causes by plotting object trajectories against checker acceptance regions, and classify the resulting bugs into task consistency, initialization, and reproducibility. Extending this audit to seven benchmarks, including RoboTwin, LIBERO-Plus, and VLABench, we identify 22 bugs of these types and 4 design limitations. For the latter, we revise permissive success checkers, correct unrealistic object masses, and add a motion-aware score that favors smoother actions. Experiments show that bug fixes can reverse method rankings, moving the baseline from last to first on one task. Addressing design limitations can likewise remove anomalous gains: on another task, the baseline moves from 21 percentage points behind an accelerated method to 5 points ahead. Gains attributed to acceleration can therefore be artifacts of the benchmark rather than better task execution. We release our bug fixes and revised benchmark settings to support trustworthy evaluation of VLA acceleration.

Qiwei Chen, Kaijun Zhou, Nuohui Shi +3

Shanghai Jiao Tong University

VLM Evaluation Efficient VLA Model Inference Vision-Language-Action Models Robot Manipulation Benchmarks

Sep 29, 2026cs.LG

Hierarchical Compression of Vision-Language Model Benchmarks

Thorough evaluation of vision-language models (VLMs) has become prohibitively expensive, as benchmarks span an ever-broader spectrum of capabilities and new models arrive at a relentless pace. Benchmark compression methods that preserve model rankings at a fraction of the cost are well studied for language models, but for VLMs the question remains under-explored. We present PRIMEBench (Pruning Redundant Items for Multimodal Evaluation), a vision-aware hierarchical benchmark compression framework that substantially reduces evaluation cost while preserving model rankings. This hierarchical framework operates in four stages: data cleaning to remove items answerable without the image and all-correct items, category representative selection to pick one benchmark per capability category, item pruning with Vision-Aware Variance (VAW), and category-count pruning. VAW combines inter-model variance with a vision-dependence score computed from multimodal embeddings alone, while encouraging coverage of diverse items within each benchmark. On models held out from item selection, it has the highest mean fidelity at the released 5% retention. The hierarchical design lets practitioners stop at any stage to match their compute budget; the released suite removes over 97% of items while preserving model rankings. Beyond compression, our analyses show how VLM evaluation behaves as model panels grow and evolve, providing guidance for designing future benchmarks that are more efficient, robust to model turnover, and explicit about the limits of evaluation-side pruning.

Hyunjong Ok, Seunggu Kang, Jaeho Lee

Pohang University of Science and Technology · Upstage AI

Sep 29, 2026cs.CV

Benchmarking Vision-Language Models on Synapse Detection and Proofreading in Connectomics

We benchmarked vision-language models (VLMs) on the decisions annotators take when inspecting electron microscopy images in connectomics: synapse detection (presence and polarity) and proofreading (split errors and merge errors). For synapse detection, we evaluated 19 open and 2 closed models across various architectures and sizes under zero-shot, four-shot in-context learning and LoRA settings, against specialist models, on datasets constructed by us using public resources. For proofreading, we evaluated 3 open and 2 closed models on the ConnectomeBench2 dataset, with cross-species transfer from fly and mouse to human and zebrafish. Most models were at chance zero-shot; a few examples helped mainly the closed and largest open ones. LoRA on a few thousand labels brought open models level with specialist models. When evaluated on unseen species, the best adapted VLMs outperformed specialist models trained on the same data in identifying merge errors. The project will be publicly available upon acceptance.

Yicong Li, Junjie Wang, Leander Lauenburg +5

Harvard University, Cambridge, MA, USA · Boston College, Chestnut Hill, MA, USA · Nanyang Technological University, Singapore

VLM Evaluation VLM Adaptation

Sep 29, 2026cs.CL

Similar Choices, Different Attention: Cross-Modal Associations in Humans and Vision-Language Models

Cross-modal associations are systematic pairings of features across modalities, such as the association of 'bouba' with round shapes and 'kiki' with sharp shapes. Prior work has compared humans and vision-language models (VLMs) on such associations, but often using different stimuli or tasks between humans and models. Here, we ask whether VLMs align with humans not only in choices, but also in where they look when making those choices. We study both VLMs and humans (N = 53), presenting them with the same stimuli, a pseudo-word and two images, and record participants' choices and eye movements, which we release. We find choice alignment in a few larger VLMs, but their saliency matches human gaze less closely than a center-bias baseline, a fixed Gaussian at the center of each image. Fine-tuning small VLMs on human choices brings their choice alignment to the level of a human majority-vote reference on unseen words and images, yet their attention still matches human gaze less closely than this baseline. Training model attention on human gaze raises attention-gaze correlation without improving choice alignment, and a single average gaze map per image position raises it by a similar amount. Matching human choices, or even human gaze patterns, is therefore not sufficient evidence of human-aligned cross-modal processing.

Sumin Hong, Katsumi Ibaraki, Renee Shi +2

University of Notre Dame

VLM Evaluation Cross-Modal Alignment Vision-Language Alignment

Sep 28, 2026cs.LG

ScAn-Bench: Evaluating Scaling Analysis Methodology

Recent progress in machine learning is driven by large-scale foundation models, where scaling laws and finding optimal scaling prescriptions for architecture, data, and hyperparameters are key in advancing the state-of-the-art. Therefore, it is surprising that no systematic study evaluates the methodology to obtain scaling laws and prescriptions across different model types. To shed light on this crucial blind spot and facilitate future research, we introduce the surrogate benchmarks ScAn-Bench-LLM and ScAn-Bench-VLM based on 4524 and 8024 checkpoints of language and vision-language model pipelines. On our benchmarks, we perform the first systematic evaluation of both data acquisition and extrapolation methodology for scaling analysis across different data modalities.

Artin Sermaxhaj, Nastaran Alipour, Donat Sinani +4

University of Freiburg · Zuse School ELIZA · Juelich Supercomputing Center (JSC), Research Center Juelich (FZJ)

VLM Evaluation LLM Evaluation Language Model Scaling Laws Scaling Laws

Sep 28, 2026cs.CL

When Words Speak Louder than Images: Towards Understanding Language Bias in Vision-Language Models

Despite substantial progress across downstream applications, vision-language models (VLMs) remain susceptible to language bias, often prioritizing linguistic cues over visual evidence and consequently producing incorrect predictions. Prior studies have proposed various approaches to understanding and mitigating language bias in VLMs, yet their findings often conflict due to the difficulty of tracing how language bias propagates within black-box VLMs. Building on the word completion task, we trace how language bias propagates through VLM inference by (1) proposing a diagnostic framework that decomposes the inference process into four distinct yet interdependent stages to trace the propagation of language bias; and (2) examining how two key factors underlying language bias, i.e., linguistic priors and cross-modal coverage, evolve across these stages and ultimately give rise to incorrect predictions. The linguistic prior captures the strength of statistical bias induced by the language model component of a VLM and represents the origin of language bias, whereas cross-modal coverage measures the extent to which linguistic cues cover the visual content. By decomposing inference into four stages and characterizing the interplay between linguistic priors and cross-modal coverage across these stages, we propose a systematic framework for tracing the propagation of language bias throughout the inference process; and uncover the underlying mechanism of language bias by revealing the interplay between linguistic priors and cross-modal coverage.

Yizhou Fang, Siyue Chen, Zimo Qi +3

University of Waterloo · Independent Researcher · Johns Hopkins University +3

VLM Evaluation Vision-Language Models VLM Interpretability Linguistic Bias in Language Models

Sep 28, 2026cs.LG

THEIA: A Multimodal Dataset and Benchmark for Vision-Language Analysis of Layout

The integration of artificial intelligence into computer-aided design frameworks has sparked a shift in the design of analog integrated circuits (ICs), transitioning the field from using manual and algorithmic-based solutions to adopting automated and intelligent paradigms. In this scenario, the GDSII file represents the industry-standard database containing the ultimate and most accurate source of information of the analog circuit, encapsulating the complex physical geometries and parasitic realities that define tape out performance. This paper proposes THEIA, a novel dataset containing thousands of layout images paired with question-answer conversations, along with a benchmark that employs a fine-tuned vision-language model (VLM) to analyze GDSII files of analog circuits, enabling designers to interact with and query physical layouts as intuitive, meaningful entities. Experimental results using thousands of analog designs across five realistic tasks demonstrate that the proposed fine-tuned VLM outperforms state-of-the-art general-purpose VLMs by a significant margin (up to 73%), highlighting a fundamental gap between general-purpose multimodal reasoning and domain-specific layout understanding.

Giuseppe Chiari, Michele Piccoli, Federico Viola +1

DEIB Politecnico di Milano Milan, Italy, 20133

VLM Evaluation Electronic Design Automation

Sep 28, 2026cs.CV

ControlTrace: Recovering Control Fields for Hidden-Content Recognition

Spatially conditioned diffusion models can embed words and contours in natural-looking images, but vision-language models (VLMs) may fail to recognize the hidden content. Transformation-based recovery depends on parameter and view selection. To evaluate hidden-content recovery and recognition, we construct FreqBlind, a 6,000-image benchmark spanning contours, real words and non-words across three conditioning strengths. The evaluated transformation-based methods show limited recognition of contour patterns and weakly conditioned hidden content. To address this limitation, we propose ControlTrace to recover the grayscale control field used during generation. An 8.4M-parameter U-Net predicts this field from the carrier image, and a VLM then identifies its content. With Qwen2.5-VL-7B-Instruct, ControlTrace achieves 60.2% open-ended contour recognition accuracy across the three conditioning strengths, exceeding the best of the three evaluated prior methods by 26.9 percentage points. On an A100 GPU, the complete pipeline adds only 7.4 ms (5.3%) to direct VLM inference. Recovered fields have lower pixel errors and higher structural similarity than the evaluated transformation views. Across four evaluated VLMs, ControlTrace retains its overall contour recognition advantage. Recognition remains stable under the tested JPEG compression, Gaussian noise and downsampling. These results support control-field recovery for hidden-content recognition in the evaluated setting.

Zijian Liu, Yaoguang Chen, Liwei Liu +4

Shanghai Jiao Tong University · Ant Group

VLM Evaluation Vision-Language Models

Sep 28, 2026cs.CV

When VLMs Trust Context: Evaluating Scene Text Recognition under Misleading Context

Vision-language models (VLMs) can read text in natural scenes, but their predictions may be influenced by the surrounding context. When the printed text conflicts with what the scene suggests, a model may return a more plausible word instead of the shown text. We introduce SceneFaith, a benchmark of 781 generated scene images for studying this behavior. Each output is classified as Literal, Canonical, or Other, separating faithful transcription from context-consistent rewriting and ordinary recognition errors. Across 15 models from seven families, all models show rewriting on clear images, with rates ranging from 8.45% to 58.51%. Controlled experiments further show that surrounding context matters: removing surrounding scene information reduces rewriting and improves literal accuracy, while changing the scene around the same text patch can also change model outputs. Moreover, weakening the target text with blur increases rewriting. These results show that reliable scene-text recognition requires VLMs to balance visual character evidence with contextual information, preserving clear text while using context mainly when the visual evidence is uncertain.

Yuxing Cheng, Yuan Wu, Yi Chang

School of Artificial Intelligence, Jilin University · Engineering Research Center of Knowledge-Driven Human-Machine Intelligence, MOE, China · International Center of Future Science, Jilin University

VLM Evaluation Counterfactual Evaluation of VLMs Scene Text Recognition Faithfulness in VLMs

Sep 28, 2026cs.AI

Before the Token Commits: Trajectory-Level Benchmarking of Visual Hallucinations in Diffusion VLMs

Multimodal diffusion language models generate responses by iteratively unmasking tokens, making each answer the endpoint of a multi-step trajectory rather than an immediate commitment. Hallucination benchmarks built for autoregressive models evaluate only the final output, and therefore cannot determine whether an unsupported claim in diffusion VLMs appears late or has already stabilized before any answer token is revealed. We introduce DynaHall, a trajectory-level benchmark of annotation-backed binary visual propositions covering object existence, counting, attributes, and relations, with controlled hard negatives graded by visual prior. DynaHall is paired with a commitment-aware protocol that records the intermediate answer tendency at every unmasking step alongside the committed output. Across five diffusion VLMs from three architecture families, visual hallucination is settled before commitment: an unsupported answer is already the preferred state while the answer position is still masked, and later unmasking steps rarely reverse it, so the failure is not introduced at the write step. This holds across decoding schedules, answer formats, and open-ended generation. DynaHall also exposes failures hidden by final-output metrics, including counting and relation collapse, prior-driven false positives, and attribute errors whose direction changes by type. Guided by this diagnosis, PGS (Pre-commitment Gradient Steering) edits still-masked answer states to reduce false positives, bringing the affirmation rate close to balance, and transfers to another architecture without degrading general ability. DynaHall and PGS suggest that hallucination should be measured and mitigated along the generation trajectory of diffusion VLMs, not only at the final answer.

Yadong Wang, Siping Yue, Yu Tian +2

Nanjing University of Aeronautics and Astronautics · State Key Laboratory of Ocean Sensing, ZJU-Hangzhou Global Scientific and Technological Innovation Center, Zhejiang University, Hangzhou, 311215, China · Security Capability Center

VLM Evaluation Vision-Language Models VLM Hallucination VLM Hallucination Mitigation

Sep 28, 2026cs.CV

When Does an Image Determine the Answer? Benchmarking Visual Answerability across Charts and Scenes

Reliable visual question answering requires correct answers when evidence is sufficient and abstention when it is not. We introduce a benchmark that connects complete-question evaluation with explicit evidence for its labels across PlotQA charts, CLEVR rendered scenes, and GQA photographs. Each question groups original and edited images, presented independently; success requires every supported answer and every required abstention to be correct. For chart missing-information labels, executable witnesses establish that admissible complete charts give different answers but identical pixels after masking. Scene labels follow source programs and edits, with a residual-cue analysis for photographs. Across 72,000 responses from six model configurations, the highest observed complete task success rates are 57.0%, 43.5%, and 33.7%, respectively. On charts, the strongest configuration achieves 96.2% per-view decision accuracy, yet 265 of its 835 groups with every decision correct still contain incorrect answers. Evaluating supported answers and necessary abstentions together exposes failures that answerability decisions alone conceal.

Sungguk Cha, Mintae Kim, Youngsub Han +2

LG Uplus, Seoul, South Korea

VLM Evaluation Visual Question Answering Chart QA LLM Abstention

Sep 27, 2026cs.CV

Seeing and Solving Are Not Enough for Vision-Language Models

Vision-language models (VLMs) answer visual questions by combining visual information extraction with downstream problem solving. We investigate a fundamental question: Does an incorrect answer necessarily reflect a failure in visual extraction or problem solving? A model may succeed at both abilities when tested separately yet still fail on the original multimodal question, a distinction that overall answer accuracy cannot reveal. To study this, we perform a question-level empirical analysis across multiple VLMs and visual domains. We define an exactly scorable task state (i.e., the visual information sufficient to solve a question) and use it to test whether the same model can extract the required state, solve the question from the ground-truth state, and answer the original multimodal question. We find that composition failures, where extraction and solving both succeed but direct answering fails, account for 17.7% to 75.6% of direct-answering errors across multiple VLMs and datasets. To address this failure mode, we introduce a simple yet effective method, termed State Realization Tuning (SRT). SRT fine-tunes LoRA adapters attached to the language-model layers while keeping the pretrained VLM weights frozen. It trains the model to output the ground-truth task state before the final answer in a single autoregressive response. SRT improves over standard supervised fine-tuning by 1.7 to 14.1 percentage points and repairs 92.5% to 98.1% of diagnosed composition failures. A single LoRA adapter trained with SRT also improves performance across substantially different task-state structures. Our work shows that having both visual extraction and problem-solving capabilities does not guarantee correct multimodal answering. Requiring the model to first output the visual information needed to solve the question can help bridge this gap.

Ziheng Wang, Mingxuan Xie, Yilin Liu +3

Sun Yat-sen University · Zhejiang University · The Hong Kong University of Science and Technology +2

VLM Evaluation Visual Question Answering Vision-Language Models VLM Reasoning

Sep 27, 2026cs.CV

A Visual Classification Dataset and Model Evaluation for Historical Manuscript Illustrations

Historical manuscript illustrations preserve rich visual evidence of past cultures. They depict people, animals, plants, diagrams, music notations, and decorative forms. Although large digitization projects have made many manuscripts available online, the material itself remains difficult to explore at scale. Extraction systems can find illustrations on manuscript pages, but without meaningful categories, large collections remain hard to search and explore. We address this gap by introducing a manually labeled dataset of 15,000 illustrations from manuscripts dating back hundreds of years across 22 categories, and evaluating modern vision models for image classification on this task. The problem is challenging due to stylistic diversity, degradation, and semantic ambiguity, with many images that fit more than one category. We compare fine-tuned CNN and Transformer-based classifiers, zero-shot CLIP, embedding-based classifiers, and direct vision-language models. Results show that fine-tuned image classifiers perform best overall, with ConvNeXt reaching 88.9% accuracy and 81.3% macro-F1. Using CLIP embeddings with XGBoost provides a strong alternative. In contrast, zero-shot CLIP and direct vision-language classification perform substantially worse, highlighting the limits of general-purpose models in this domain. Beyond overall performance, the analysis reveals which categories are visually separable and where errors reflect genuine semantic overlap, suggesting that some limitations arise from the taxonomy itself.

Yoav Evron, Michal Bar-Asher Siegal, Michael Fire

Faculty of Computer and Information Science, Ben-Gurion University of the Negev, Be’er Sheva, Israel · The Goldstein-Goren Department of Jewish Thought, Ben-Gurion University of the Negev, Be’er Sheva, Israel

VLM Evaluation Historical Document Analysis Zero-Shot Image Classification Image Classification Historical Document Processing

Sep 27, 2026cs.LG

MoGround: Measuring and Mitigating Modality Distraction in Vision-Language Models

We release MoGround, a vision-language dataset spanning four visual domains in which the answer to every question is guaranteed to be available from exactly one modality. This guarantee enables us to measure modality distraction, the failure in which a model answers a question correctly from one modality alone and then flips to a wrong answer once irrelevant content from the other modality is added. Existing probes rarely establish single-modality answerability this way, making it hard to isolate distraction in the first place. Across seven open-source VLMs, we find that modality distraction is not universal but model-dependent. The weaker-grounded modality is the more distracted one (r = +0.86), and distraction scales inversely with grounding strength (r = -0.90). The single-modality guarantee also enables a mitigation method that needs to distinguish between relevant and irrelevant context. Trained on one split of MoGround alone, a weight-space robustness vector reduces distraction on all seven models by 9% to 51%, at a cost of only 0.1 average points of accuracy on standard multimodal tasks.

Luca Zhou, Bo Zhao, Rose Yu +2

Sapienza University of Rome · Harvard University · UC San Diego +2

VLM Evaluation Multimodal Grounding VLM Robustness

Sep 26, 2026cs.CV

EyeVQA: Benchmarking Ophthalmic Vision-Language Models from Recognition to Spatial Grounding

Vision-language models (VLMs) have shown increasing potential for medical image understanding, yet their capabilities in ophthalmic imaging remain insufficiently characterized. Existing ophthalmic datasets are typically designed for individual diseases or specialized tasks, making it difficult to systematically evaluate whether VLMs can move beyond disease recognition toward comparative reasoning and fine-grained spatial grounding. We introduce EyeVQA, a unified visual question answering benchmark for comprehensive evaluation of ophthalmic VLMs. EyeVQA is constructed from 21 available ophthalmic datasets and contains 20,000 question-answer pairs spanning six disease groups and seven question types: Single-Choice, Multi-Select, Variable-Select, True-False, Ranking, Point Location, and Bounding Box. Gold answers are deterministically derived from source-provided diagnoses, severity grades, clinical findings, segmentation masks, bounding boxes, and anatomical landmarks, enabling reproducible evaluation without relying on model-generated annotations. Notably, 44.5% of the questions require reasoning across multiple images, extending evaluation beyond conventional single-image medical VQA. We benchmark fourteen representative general-purpose, scientific, and medically specialized VLMs under a unified zero-shot protocol. The best-performing model only achieves an overall score of 62.8, while substantial gaps remain in spatial grounding and cross-task generalization. These results highlight the limitations of current VLMs in comprehensive ophthalmic visual understanding and establish EyeVQA as a diagnostic benchmark for developing more reliable and spatially grounded ophthalmic multimodal models. The project page is available at https://github.com/PKUTHM/EyeVQA.

Gujie Shao, Zixun Xie, Xuechun Xing +7

VLM Evaluation Medical VQA Vision-Language Grounding Medical Image Benchmarks

Sep 24, 2026cs.CV

GHOST-Q: Towards Studying Grounding Hallucinations Overlooked Under Same-score TradeOffs in Quantized VLMS

Post-training quantization of vision--language models (VLMs) is typically assessed through aggregate task accuracy and memory savings, but preserving a headline score does not guarantee preservation of visual grounding behavior. We present GHOST-Q, a cross-precision controlled evaluation of three 8B VLM families under FP16, INT8, and NF4 across utility and hallucination-sensitive benchmarks. Rather than comparing only aggregate accuracy, we pair FP16 and quantized predictions item by-item to quantify how compression redistributes grounding successes and failures. Five of six quantized variants preserve MMStar accuracy within

\pm2

percentage points, yet 10 of 36 paired effects remain significant after false-discovery-rate correction, nine on hallucination-sensitive conditions. Same-device A100 profiling further demonstrates that substantial memory reduction does not necessarily mean lower inference latency. Finally, an open-ended AMBER audit reveals strong generation budget censoring whose severity varies by architecture and precision. These results show that quantized VLMs should be evaluated jointly for aggregate utility, grounding reliability, generation behavior, and realized deployment efficiency.

Saim Rehman, Muhammad Shafique

eBRAIN Lab, Division of Engineering, New York University Abu Dhabi (NYUAD), Abu Dhabi, UAE

VLM Evaluation VLM Quantization Vision-Language Grounding Post-Training Quantization

Sep 24, 2026cs.CL

An Empirical Study of VLM Pipelines for Long-Document QA

Vision-Language Models (VLMs) are increasingly used for long-document processing, where the inputs combine text with charts, tables, figures, and complex layouts. Deploying them means choosing how to feed the document to the model, which retriever to use when only a subset of pages is sent, and whether to run the model agentically or as a static pipeline. We study these choices on two long-document QA benchmarks with both frontier API and open-weight VLMs. First, on MMLongBench-Doc our six-tool agent with page, table, figure, and search calls pays off only once the answering VLM is large enough: with Qwen3.5-4B and 9B it trails static page input, with Qwen3.5-27B it draws level, and with Sonnet 4.5 it leads. On LongDocURL it is level with or ahead of static input at every reader. Its lead over the strongest static pipeline is clearest with the frontier reader on MMLongBench-Doc and narrows to within noise on LongDocURL. Second, retrieval modality matters more than the specific retriever: the strongest image retriever leads the strongest text pipeline, and on the text side a single off-the-shelf cross-encoder rerank essentially matches a much heavier multi-stage LLM pipeline. Top-k image retrieval is also the most token-efficient input at every reader we paired it with, at roughly a seventh to a quarter of the tokens of sending every page. Third, cutting across all three choices, three of our strongest pipelines succeed on different questions, and an oracle that picks the best pipeline per question gains roughly thirteen points over the best single pipeline, though evidence-type routing recovers almost none of it.

Kenan E. Ak, Jay Mohta, Gwang Gook Lee +2

VLM Evaluation Tool Use in VLMs Document QA Visual Document Retrieval Long-Context VLMs

Sep 24, 2026cs.CL

Reasoning Instructions Can Break Answer Decoding in Vision--Language Models

Chain-of-thought (CoT) instructions can distort multiple-choice VLM evaluation when a scorer appends a reasoning cue but reads answer-label logits before the model generates any rationale. We call this CoT-prefix scoring. On ScienceQA, Qwen2.5-VL-7B drops from 80.76% to 45.48%, and across five option-content permutations 93.54% of CoT-prefix predictions select the first slot. Condition-matched linear probes recover 78.94% from the same hidden states, while free generation restores 75.24%, showing that the answer often survives the prefix and the immediate readout fails. Vocabulary and layer diagnostics explain the mismatch: probability mass moves toward continuation tokens, while answer information remains linearly accessible in late layers. The effect recurs with varying severity across datasets and models, though not universally. These results show that CoT-prefix scoring can confound model knowledge with an evaluation-interface mismatch and should be avoided unless the requested and scored output events are aligned.

Zeyan Li, Siyuan Qiu, Jianfeng Xu

Shanghai Jiao Tong University

VLM Evaluation Language Model Generation Evaluation Multiple-Choice Question Answering CoT Reasoning

Sep 23, 2026cs.CV

Gender Bias in Vision-Language In-Context Learning

In-context learning (ICL) enables large vision-language models (LVLMs) to perform tasks by following patterns from in-context examples, yet its potential to amplify societal biases remains underexplored. We systematically investigate how ICL influences gender bias in LVLMs through VL-BICLE, an evaluation framework comprising six ICL settings, three tasks, and four datasets. Our experiments on six LVLMs reveal that gendered ICL demonstrations act as a directional force, shifting model bias toward the demonstrated gender through a cross-gender mechanism that disproportionately degrades performance on the opposite gender. This effect appears in image captioning and pronoun prediction but not in visual question answering, indicating that gendered ICL influences bias only when the task output involves gendered language. Similarity-based retrieval methods inherit the training pool's gender imbalance and offer no debiasing advantage, while standard quality metrics remain blind to these bias shifts. To mitigate this bias, we replace real in-context images with synthetic ones from stable diffusion models while keeping captions unchanged. This simple intervention reduces gender bias without degrading caption quality.

Tong Xiang, Yuta Nakashima, Noa Garcia

The University of Osaka, Japan

VLM Evaluation In-Context Learning

Sep 23, 2026cs.CV

Beyond Balanced Accuracy: A Resolution and Parity-Controlled Benchmark for Vision-Language and Vision-Only Defect Assessment in UAV Power-Line Inspection

Vision-language models (VLMs) are often reported to outperform task-specific vision backbones for unmanned aerial vehicle (UAV) power-line defect assessment. We test that claim on ElecVQA-Bench, a 56,972-item benchmark derived from the public InsPLAD dataset, across six evaluation choices: partition, evaluated item set, label space, replication, input resolution, and side information. On a matched partition, a Swin Transformer and the strongest adapted VLM differ by only 0.03 points at binary screening. At seven-way defect typing, increasing the vision backbones from 224 px to the measured pixel budget of the VLM preprocessor narrows the gap against InternVL3.5-8B from +20.53 to -0.57 points for ResNet-50 and from +23.67 to +4.70 points for Swin-T. A pixel-budget audit shifts Qwen3-VL-8B macro recall by 10.78 points, yet a source-pixel-matched InternVL control still leaves Qwen ahead by 7.43 to 13.61 points while using 56% fewer visual tokens, so neither source pixels nor token budget explains the difference between the two VLMs. A two-seed global replication changes Qwen binary accuracy and seven-way macro recall by 0.86 and 1.02 points. After split-specific retraining, Qwen does not lead at crop or image level, and a 14-tower, three-seed replication reverses the sign across seeds, giving mean common-six macro recall of 0.9085 for Qwen against 0.9509 for ResNet-50. No split regime yields a family-level advantage that survives multiple-comparison correction. The study supports a benchmark-audit contribution rather than a general claim of VLM superiority.

Linghao Zhang, Siyu Xiang, Junwei Kuang +1

State Grid Sichuan Electric Power Research Institute, Chengdu 610041, China · Power System Security and Operation Key Laboratory of Sichuan Province

VLM Evaluation Anomaly Detection

Sep 23, 2026cs.CV

What Looks Like a Capability Limit in Vision-Language Models Is a Readout Limit

Benchmarks for vision-language models offer their answer choices in some convention: a letter, a color name, a pixel coordinate. That convention is treated as neutral. We find it is not, and that the limits a benchmark reports can belong to the readout rather than to the model. On 200 COCO photographs, Qwen3-VL-4B picks the correct one of nine locations for a named object 68.5% of the time when the locations are given in English and 20.0% when the same locations are given as pixel coordinates. Chance is 11.1%. The cost arises when the answer options are coordinates; giving the model a coordinate in the question instead costs 3.5 points and is not significant. The gap holds on a 4x4 grid, under 8-bit rather than 4-bit quantization, and in every slice by object size, boundary distance and category. It also decides which model wins. Two models that tie under English names differ by 39 points in one coordinate system and by 54 in the other, in opposite directions. On the color task, three of the four open models capable of the task show the penalty; on photographs, two of three open models do, and so does Gemini, at 11.1 points on parseable answers (p = 1e-4). GPT-4o does not. To ask whether a model reads a coordinate at all, we attach the wrong name to each one and record which the model follows. Color options written as hue angles are followed below chance; a normalized pixel convention is followed at four times chance. This tells apart conventions a model can use from ones it cannot, though it did not predict accuracy on two untried conventions. Five models also name the same color wheel five different ways, so a fixed answer vocabulary is not neutral across models either. Five times during this work we measured a capable model as incapable because our scorer and the model disagreed about what an answer looks like. We report each case. They are the phenomenon in miniature.

Alfredo F. Frontera Del Valle

Columbia University

VLM Evaluation Vision-Language Grounding

Sep 23, 2026cs.CL

PRISM-VLM: A Multi-Axis Discriminative Benchmark for Compact Vision-Language Models

Compact vision-language models (VLMs) now power a growing share of multimodal applications. The benchmarks used to compare them, however, inherit a frontier-centric design: each model is reduced to a single accuracy number, narrowing the inter-model gap on saturated suites and pressing models into low-score bands on harder ones. We introduce PRISM-VLM, a multi-axis discriminative benchmark that scores every item along seven axes covering the recurring failure modes (task quality, behavioral robustness, and capability bottlenecks) and combines them into a single PScore, with items recycled from fifteen public benchmarks. Across compact VLMs from the past two years, PScore separates model pairs more reliably than prior single-axis benchmarks under an item-level paired bootstrap, and surfaces behavioral differences these benchmarks average away. Even models with statistically indistinguishable PScores diverge sharply along the per-axis profile, particularly on sycophancy, which is nearly orthogonal to single-prompt accuracy. We will release the full pipeline, prompts, and per-item annotations.

Sanghee Park, Kee-Eung Kim

NAVER Cloud AI · KAIST AI

VLM Evaluation VLM Robustness

Sep 22, 2026cs.CV

Feed the Panel Dimensions, Not Verdicts: Rubric-Decomposed Fusion of Vision-Language Aesthetic Judges

Vision-language models (VLMs) are deployed as zero-shot judges of image aesthetics, and panels of several models are recommended, on thin evidence, as the way to make such judges reliable. On two human-rated datasets, EVA and PARA, we find that a panel of holistic judges never significantly beats its best member, whether the verdicts are averaged or fused by a learned combiner. What a panel is worth depends on what it is fed. We therefore have each model score each image on the five dimensions of a frozen, human-written rubric and fuse those scores, alongside each model's verdict, across model families with an out-of-fold combiner. The dimension scores measure what their labels claim: with the overall human score partialled out, a dimension prompt carries more attribute-specific information than the holistic prompt in 28 of 30 model-attribute cells. Fused, they beat the best single VLM in all ten three-family panels on EVA (against that best single model, +0.07 Spearman rho for the strongest trio and +0.10 for the pre-declared one, and +0.06 and +0.07 when averaged over twenty fold partitions; against the panel mean, the primary test gives +0.118 on its EVA design set), and on PARA they reach parity under Spearman rho and a small, non-significant loss under Kendall tau-b, where one model already captures 85% of the human noise ceiling. It is not a feature-count artefact: giving the same combiner an equal number of pure holistic columns, split from the same repetitions, does not reproduce it. The gain costs a few hundred labels, which do not transfer between datasets, and 4.8x the API calls on EVA; we report it with paired bootstraps and Kendall tau-b, alongside a failed pre-registration and the configurations that lost.

Amit Jadhav, Shaurya Beriwala, Beomjin Kim

Purdue University Fort Wayne

VLM Evaluation Vision-Language Models Rubric-Based Evaluation Vision-Language Model-as-a-Judge

Sep 21, 2026cs.CV

Do LiDAR Language Models Really Understand Spatio-temporal Relationships?

Recent 4D LiDAR language models aim to reason about objects and their evolving spatial relationships. Yet, in our evaluation, always selecting the same option nearly matches the multiple-choice accuracy of two B4DL-derived configurations. We introduce LiDAR-Hallu, a geometry-referenced benchmark and diagnostic protocol with 10,000 questions across 150 nuScenes scenes. It covers object existence, ego-relative position, distance ordering, relative motion, and temporal localization, with explicit rules for selecting objects, comparing times, and determining reference answers. Our protocol combines fixed-answer and candidate-content controls, cross-scene pairs with identical prompts but opposite reference answers, and relation-specific recall. Analysis of 100,000 recorded responses reveals failures hidden by aggregate accuracy. Candidate duration alone makes temporal answers predictable without observing LiDAR. On paired questions, the models frequently give the same answer to scenes requiring opposite answers. Relation-specific analysis further shows that both configurations miss every positive lateral-motion case across all tested conditions. Temporal-shuffle contrastive decoding provides little net improvement, as repairs are largely offset by new errors and the main failures persist. These results show that evaluating spatio-temporal reasoning requires testing whether models distinguish the queried physical relationships, rather than relying on individual-answer accuracy alone. The source code, checkpoints, and data are released at https://github.com/Awesome4D/4DMLLM_Hallucination_Bench.

Runyi Yang, Murat Akkoyun, Di Wen +8

INSAIT, Sofia University “St. Kliment Ohridski”, Bulgaria · Karlsruhe Institute of Technology, Germany · Hunan University, China

VLM Evaluation Hallucination Detection in VLMs Spatial Reasoning Benchmarks Spatiotemporal Reasoning 3D VLMs

Sep 21, 2026cs.CV

ChartJudgeBench: Evaluating LMM Judges for Chart-to-Code Generation

Building strong chart-to-code systems increasingly relies on reinforcement learning, whose effectiveness depends critically on the quality of the reward signal. Large Multimodal Models (LMMs) play a natural critical role in jointly assessing chart visual appearance and task requirements. They are therefore increasingly used as visual critics and reward models, yet their reliability as judges remains largely unexplored. To this end, we introduce ChartJudgeBench, a diagnostic vision-language benchmark for assessing LMM judges in chart-to-code workflows. It includes 1,003 Chart Perception Alignment (CPA) instances for pairwise chart comparison and 650 Chart Reasoning Judgment (CRJ) instances for binary Accept/Reject verification in Chart Reproduction and Chart Editing. Together, these tasks emulate the core judging decisions required in agentic refinement and RL-based chart optimization. Our evaluation of strong LMMs reveals four systematic limitations: (i) positional bias in pairwise comparison, (ii) a strong tendency to overpredict Accept, (iii) difficulty in matching visual styles and aesthetics, and (iv) an unexpected leniency bias in RL-trained models. These findings show that current LMM judges require explicit reliability validation before being used as critics or reward models in chart-to-code optimization. The code and data are available on ChartJudgeBench.

Lijian Wu, Henry Hengyuan Zhao, Zijian Zhang +3

CSU-JPG, Central South University · National University of Singapore

VLM Evaluation LLM Evaluation LLM-as-a-Judge Code Generation LLM Reliability Vision-Language Model-as-a-Judge

Sep 21, 2026cs.CV

LegendBench: A Diagnostic Benchmark for Legend Understanding with Counterfactual Interventions

Legends are fundamental to chart understanding, as reliable interpretation requires correctly binding legend entries to corresponding visual marks. While vision-language models (VLMs) are increasingly applied to chart understanding, their legend understanding is poorly diagnosed by aggregate accuracy, which can be satisfied by superficial shortcuts and confound legend-specific errors with other reasoning failures. To enable fine-grained diagnosis and controlled testing, we introduce LegendBench, a parametric benchmark and generation pipeline that produces targeted legend-centric test cases. LegendBench contributes (1) a capability-task taxonomy spanning legend parsing, legend grounding, legend-conditioned reasoning, and legend-aware abstention to localize failures, and (2) counterfactual group generation, where each base chart yields multiple variants under controlled legend interventions to probe model invariance and sensitivity. Using LegendBench, we evaluate both general-purpose VLMs and specialized chart models and generate their capability profiles, revealing persistent bottlenecks in reliable legend-to-mark binding and counterfactual consistency. We then use these capability profiles to guide targeted fine-tuning, demonstrating that bottleneck-specific interventions can effectively close the localized capability gaps and generalize to unseen data. We further leverage our counterfactual design to conduct fine-grained diagnostic experiments, analyzing encoding-channel effects, legend-order shortcuts, and abstention under varying visibility.

Xinnuo Zhang, Zhike Tang, Jing Xu +2

The Hong Kong University of Science and Technology (Guangzhou)

VLM Evaluation Counterfactual Evaluation of VLMs Chart Understanding

Sep 17, 2026cs.CV

Benchmarking MLLMs via Cognitive Expected Scene Graph for Safety-Critical Visual Negation Understanding

True machine intelligence requires transcending passive pixel registration to master top-down functional reasoning over absent information via visual negation understanding. However, unconstrained visual negation paradigms remain overly open-ended, and pervasive affirmation bias causes both existing Multi-Modal Large Language Models (MLLMs) and evaluation metrics to fail under negative semantics. To solve these intertwined challenges systematically, we first anchor the boundaries of negation reasoning within specific cognitive goals. Specifically, by focusing on safety as a highly pragmatic and critical cognitive dimension, we define the task of \textbf{S}cene \textbf{N}egation \textbf{U}nderstanding under \textbf{S}afety Cognition (\textbf{SNUS}). Under this framework, we construct a high-fidelity negative caption dataset mapping dense assertions of localized hazards. Concurrently, we propose the Cognitive Expected Scene Graph (CESG) Score, a structure-grounded, polarity-aware evaluation metric. Extensive experiments demonstrate that while current models struggle on the task, traditional metrics completely collapse under semantic reversals. Conversely, our framework delivers a solid benchmark for SNUS, providing a rigorous foundation to advance risk-aware situational comprehension and counterfactual cognition.

Zhiyun Jiang, Hanyong Wang, Binbin Liang +3

Sichuan University, Chengdu, China · Beijing Institute of Technology, Beijing, China

Sep 16, 2026cs.CV

Can Vision-Language Models Judge Olympic Diving? From Reasoning to Scores in Zero-Shot Action Quality Assessment

Automated action quality assessment (AQA) in Olympic sports remains a challenging task due to the complexity of human motion and the subjectivity inherent in expert judging. This work evaluates the capability of open-source Vision-Language Models (VLMs) to perform zero-shot action quality assessment on Olympic diving videos using the AQA-7 benchmark dataset. In this regard, a regression-based framework is pro-posed to leverage both the semantic reasoning and phase-level sub-scores generated by the VLMs, combining TF-IDF vectorization, dimensionality reduction, and ensemble learning to predict final competition scores. Experimental results show that standalone VLMs achieve moderate Spearman correlations below 0.32, while the proposed ensemble regression framework substantially improves performance in the reported evaluation, reaching a Spearman correlation of 0.67 with a four-model configuration. Textual reasoning features con-sistently outperformed raw numerical sub-scores, highlighting the richness of VLM-generated explanations for action quality analysis. These findings suggest that VLMs hold strong potential as assistive tools for explainable and semi-automated sports performance evaluation. The code is publicly available on GitHub https://github.com/hvelesaca/olympic diving judge vlm

Henry O. Velesaca, David Freire-Obregon, Luigi Miranda +1

ESPOL Polytechnic University, Campus Gustavo Galindo, Guayaquil, Ecuador · Software Engineering Department, Research Center for Information and Communication Technologies (CITIC-UGR), University of Granada, 18071, Granada, Spain · SIANI, Universidad de Las Palmas de Gran Canaria, Spain +1

VLM Evaluation Action Quality Assessment