Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs
Open accessAbstract
Is vision good enough for language? Recent advancements in multimodal models primarily stem from the powerful reasoning abilities of large language models (LLMs). However, the visual component typically depends only on the instance-level contrastive language-image pre-training (CLIP). Our research reveals that the visual capabilities in recent MultiModal LLMs (MLLMs) still exhibit systematic shortcomings. To understand the roots of these errors, we explore the gap between the visual embedding space of CLIP and vision-only self-supervised learning. We identify “CLIP-blind pairs”- images that CLIP perceives as similar despite their clear visual differences. With these pairs, we construct the Multimodal Visual Patterns (MMVP) benchmark. MMVP exposes areas where state-of-the-art systems, including GPT-4V, struggle with straightforward questions across nine basic visual patterns, often providing incorrect answers and hallucinated explanations. We further evaluate various CLIP-based vision-and-language models and found a notable correlation between visual patterns that challenge CLIP models and those problematic for multimodal LLMs. As an initial effort to address these issues, we propose a Mixture of Features (MoF) approach, demonstrating that integrating vision self-supervised learning features with MLLMs can significantly enhance their visual grounding capabilities. Together, our research suggests visual representation learning remains an open challenge, and accurate visual grounding is crucial for future successful multimodal systems.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Recent studies have identified key limitations of VLMs that align with our findings, including biases toward natural image-text pairs, optimization for image description rather than action-oriented reasoning, and challenges with out-of-distribution inputs.
“Recent studies have identified key limitations of VLMs that align with our findings, including biases toward natural image-text pairs, optimization for image description rather than action-oriented reasoning, and challenges with out-of-distribution inputs (Tan et al., 2024; Tong et al., 2024; Rahmanzadehgervi et al., 2024; Zang et al., 2024; Guan et al., 2023).”
From BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games (2024) · cites this paper as Tong et al., 2024
Recent studies highlight a number of shortcomings of frontier multimodal models.
“Recent studies highlight a number of shortcomings of frontier multimodal models (Moskvichev et al., 2023; Tong et al., 2024; Chen et al., 2024a; Fu et al., 2024b).”
From Does Spatial Cognition Emerge in Frontier Models? (2024) · cites this paper as Tong et al., 2024
One shortcoming of frontier multimodal models is that they may not perceive the image in detail, often missing fine-grained details or ignoring the image entirely.
“One such shortcoming is that models may not perceive the image in detail, often missing fine-grained details or ignoring the image entirely (Chen et al., 2024b; Guan et al., 2024; Tong et al., 2024).”
From Does Spatial Cognition Emerge in Frontier Models? (2024) · cites this paper as Tong et al., 2024
Recent literature indicates that vision language models are primarily bottlenecked by their vision module.
“This is consistent with recent findings in the literature that vision language models are primarily bottlenecked by their vision module49,50.”
From Generating dermatopathology reports from gigapixel whole slide images with HistoGPT (2025) · cites this paper as [49]
Cites
ImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380 · see citing works
High-Resolution Image Synthesis with Latent Diffusion Models
Rombach, R., Blattmann, A., Lorenz, D., Esser, P., & Ommer, B. (2022) · 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52688.2022.01042 · Cited by 15,880 · see citing works
Momentum Contrast for Unsupervised Visual Representation Learning
He, K., Fan, H., Wu, Y., Xie, S., & Girshick, R. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr42600.2020.00975 · Cited by 13,347 · see citing works
Self-citationExploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., Zhou, Y., Li, W., & Liu, P. J. (2019) · arXiv · doi:10.48550/arxiv.1910.10683 · Cited by 10,929 · see citing works
Masked Autoencoders Are Scalable Vision Learners
He, K., Chen, X., Xie, S., Li, Y., Dollar, P., & Girshick, R. (2022) · 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52688.2022.01553 · Cited by 8,489 · cited 1× · see citing works
“This category includes contrastive SSL [5, 7, 16, 17] and mask-based SSL [4, 18, 70].”
BackgroundSelf-citationVisual Genome: Connecting Language and Vision Using Crowdsourced Dense Image Annotations
Krishna, R., Zhu, Y., Groth, O., Johnson, J., Hata, K., Kravitz, J., Chen, S., Kalantidis, Y., Li, L.-J., Shamma, D. A., et al. (2017) · International Journal of Computer Vision · doi:10.1007/s11263-016-0981-7 · Cited by 5,107 · see citing works
LLaMA: Open and Efficient Foundation Language Models
Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M.-A., Lacroix, T., Rozière, B., Goyal, N., Hambro, E., Azhar, F., et al. (2023) · arXiv · doi:10.48550/arxiv.2302.13971 · Cited by 4,602 · see citing works
InstructBLIP: Towards General-purpose Vision-Language Models with Instruction TuningPreprint
Dai, W., Li, J., Li, D., Tiong, A., Zhao, J.-Q., Wang, W., Li, B. A., Fung, P., & Hoi, S. (2023) · Neural Information Processing Systems · doi:10.48550/arxiv.2305.06500 · Cited by 4,191 · see citing works
GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering
Hudson, D. A., & Manning, C. D. (2019) · Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2019.00686 · Cited by 3,703 · see citing works
Llama 2: Open Foundation and Fine-Tuned Chat Models
Touvron, H., Martin, L., Stone, K., Albert, P., Almahairi, A., Babaei, Y., Bashlykov, N., Batra, S., Bhargava, P., Bhosale, S., et al. (2023) · arXiv · doi:10.48550/arxiv.2307.09288 · Cited by 2,604 · see citing works
Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering
Goyal, Y., Khot, T., Summers-Stay, D., Batra, D., & Parikh, D. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.670 · Cited by 2,287 · see citing works
Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena
Zheng, L., Chiang, W.-L., Sheng, Y., Zhuang, S., Wu, Z., Zhuang, Y., Lin, Z., Li, Z., Li, D., Xing, E. P., et al. (2023) · arXiv · doi:10.48550/arxiv.2306.05685 · Cited by 1,888 · see citing works
Conceptual Captions: A Cleaned, Hypernymed, Image Alt-text Dataset For Automatic Image Captioning
Sharma, P., Ding, N., Goodman, S., & Soricut, R. (2018) · Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/p18-1238 · Cited by 1,879 · see citing works
Improved Baselines with Visual Instruction Tuning
Liu, H., Li, C., Li, Y., & Lee, Y. J. (2024) · 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52733.2024.02484 · Cited by 1,748 · see citing works
Sparks of Artificial General Intelligence: Early experiments with GPT-4
Bubeck, S., Chandrasekaran, V., Eldan, R., Gehrke, J., Horvitz, E., Kamar, E., Lee, P., Lee, Y. T., Li, Y., Lundberg, S., et al. (2023) · arXiv · doi:10.48550/arxiv.2303.12712 · Cited by 1,526 · see citing works
Generation and Comprehension of Unambiguous Object Descriptions
Mao, J., Huang, J., Toshev, A., Camburu, O., Yuille, A., & Murphy, K. P. (2016) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr.2016.9 · Cited by 1,211 · see citing works
Sigmoid Loss for Language Image Pre-Training
Zhai, X., Mustafa, B., Kolesnikov, A., & Beyer, L. (2023) · 2023 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01100 · Cited by 1,127 · see citing works
ReferItGame: Referring to Objects in Photographs of Natural Scenes
Kazemzadeh, S., Ordóñez, V., Matten, M., & Berg, T. L. (2014) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.3115/v1/d14-1086 · Cited by 1,082 · see citing works
DINOv2: Learning Robust Visual Features without Supervision
Oquab, M., Darcet, T., Moutakanni, T., Vo, H., Szafraniec, M., Khalidov, V., Fernandez, P., Haziza, D., Massa, F., El-Nouby, A., et al. (2023) · arXiv · doi:10.48550/arxiv.2304.07193 · Cited by 1,078 · cited 7× · see citing works
“Our exploration includes the use of two visual encoders: CLIP-ViT-L-14 [43] and DINOV2-ViT-L-14 [42].”
BackgroundMethodInfluentialThe Dawn of LMMs: Preliminary Explorations with GPT-4V(ision)Preprint
Yang, Z.-Y., Li, L., Lin, K., Wang, J., Lin, C.-C., Liu, Z., & Wang, L.-J. (2023) · arXiv.org · doi:10.48550/arxiv.2309.17421 · Cited by 1,036 · see citing works
Self-citationTowards VQA Models That Can Read
Singh, A., Natarajan, V., Shah, M., Jiang, Y. L., Chen, X., Batra, D., Parikh, D., & Rohrbach, M. (2019) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr.2019.00851 · Cited by 777 · see citing works
OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge
Marino, K., Rastegari, M., Farhadi, A., & Mottaghi, R. (2019) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr.2019.00331 · Cited by 749 · see citing works
OCR-VQA: Visual Question Answering by Reading Text in Images
Mishra, A., Shekhar, S., Singh, A., & Chakraborty, A. (2019) · IEEE International Conference on Document Analysis and Recognition · doi:10.1109/icdar.2019.00156 · Cited by 746 · see citing works
High-Resolution Image Synthesis with Latent Diffusion Models
Rombach, R., Blattmann, A., Lorenz, D., Esser, P., & Ommer, B. (2021) · arXiv · doi:10.48550/arxiv.2112.10752 · Cited by 737 · see citing works
VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised LearningPreprint
Bardes, A., Ponce, J., & LeCun, Y. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2105.04906 · Cited by 617 · see citing works
Self-citationEvaluating Object Hallucination in Large Vision-Language Models
Li, Y., Du, Y., Zhou, K., Wang, J., Xin, Z., & Wen, J.-R. (2023) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2023.emnlp-main.20 · Cited by 580 · see citing works
Mitigating Gender Bias in Natural Language Processing: Literature Review
Sun, T., Gaut, A., Tang, S., Huang, Y., ElSherief, M., Zhao, J., Mirza, D., Belding, E., Chang, K.-W., & Wang, W. Y. (2019) · Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/p19-1159 · Cited by 492 · see citing works
MMBench: Is Your Multi-modal Model an All-around Player?
Liu, Y., Duan, H., Zhang, Y., Li, B., Zhang, S., Zhao, W., Yuan, Y., Wang, J., He, C., Liu, Z., et al. (2023) · arXiv · doi:10.48550/arxiv.2307.06281 · Cited by 487 · see citing works
Self-citationMiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models
Zhu, D., Chen, J., Shen, X., Li, X., & Elhoseiny, M. (2023) · arXiv · doi:10.48550/arxiv.2304.10592 · Cited by 469 · see citing works
On Measuring Social Biases in Sentence Encoders
May, C., Wang, A., Bordia, S., Bowman, S. R., & Rudinger, R. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1063 · Cited by 461 · see citing works
Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture
Assran, M., Duval, Q., Misra, I., Bojanowski, P., Vincent, P., Rabbat, M., LeCun, Y., & Ballas, N. (2023) · 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52729.2023.01499 · Cited by 456 · see citing works
Self-citationGonen, H., & Goldberg, Y. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1061 · Cited by 454 · see citing works
EVA-CLIP: Improved Training Techniques for CLIP at ScalePreprint
Sun, Q., Fang, Y., Wu, L., Wang, X., & Cao, Y. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2303.15389 · Cited by 299 · see citing works
A-OKVQA: A Benchmark for Visual Question Answering Using World Knowledge
Schwenk, D., Khandelwal, A., Clark, C., Marino, K., & Mottaghi, R. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20074-8_9 · Cited by 295 · see citing works
Data Filtering NetworksPreprint
Fang, A., Jose, A. M., Jain, A., et al. (2023) · International Conference on Learning Representations · doi:10.48550/arxiv.2309.17425 · Cited by 292 · see citing works
SLIP: Self-supervision Meets Language-Image Pre-training
Mu, N., Kirillov, A., Wagner, D., & Xie, S. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19809-0_30 · Cited by 285 · see citing works
Self-citationTextCaps: A Dataset for Image Captioning with Reading Comprehension
Sidorov, O., Hu, R., Rohrbach, M., & Singh, A. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-58536-5_44 · Cited by 242 · see citing works
Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality
Thrush, T., Jiang, R., Bartolo, M., Singh, A., Williams, A., Kiela, D., & Ross, C. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00517 · Cited by 240 · see citing works
LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init AttentionPreprint
Zhang, R., Han, J., Liu, C., Gao, P., Zhou, A., Hu, X., Yan, S., Lu, P., Li, H., & Qiao, Y. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2303.16199 · Cited by 202 · see citing works
MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models
Fu, C., Chen, P., Shen, Y., Qin, Y., Zhang, M., Xu, L., Yang, J., Zheng, X., Li, K., Sun, X. W., et al. (2025) · neural information processing systems · doi:10.52202/085713-4899 · Cited by 148 · see citing works
Shap-E: Generating Conditional 3D Implicit FunctionsPreprint
Jun, H., & Nichol, A. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2305.02463 · Cited by 115 · see citing works
MM-Vet: Evaluating Large Multimodal Models for Integrated CapabilitiesPreprint
Yu, W., Yang, Z., Li, L., Wang, J., Lin, K., Liu, Z., Wang, X., & Wang, L. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2308.02490 · Cited by 74 · see citing works
Self-citationNotebaard, J. (2024) · doi:10.33540/2168 · Cited by 57 · see citing works
The effectiveness of MAE pre-pretraining for billion-scale pretraining
Singh, M., Duval, Q., Alwala, K. V., Fan, H., Aggarwal, V., Adcock, A., Joulin, A., Dollár, P., Feichtenhofer, C., Girshick, R., et al. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00505 · Cited by 56 · see citing works
Xu, H., Xie, S., Tan, X. E., Huang, P.-Y., Howes, R., Sharma, V., Li, S.-W., Ghosh, G., Zettlemoyer, L., & Feichtenhofer, C. (2023) · arXiv · doi:10.48550/arxiv.2309.16671 · Cited by 34 · see citing works
Self-citationInvestigating the Catastrophic Forgetting in Multimodal Large Language Models
Zhai, Y., Tong, S., Li, X., Cai, M., Qu, Q., Lee, Y. J., & Ma, Y. (2023) · arXiv · doi:10.48550/arxiv.2309.10313 · Cited by 24 · see citing works
Self-citationImage Captioners Are Scalable Vision Learners TooPreprint
Tschannen, M., Kumar, M., Steiner, A., et al. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2306.07915 · Cited by 10 · see citing works
ConvNet vs Transformer, Supervised vs CLIP: Beyond ImageNet AccuracyPreprint
Vishniakov, K., Shen, Z., & Liu, Z. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2311.09215 · Cited by 1 · see citing works
Self-citation
Flagged first, then by citations
Cited by
Sorted by citations
Hallucination of Multimodal Large Language Models: A Survey
Bai, Z., Wang, P., Xiao, T., He, T., Han, Z., Zhang, Z., & Shou, M. Z. (2024) · arXiv · doi:10.48550/arxiv.2404.18930 · Cited by 38 · see citing works
Generating dermatopathology reports from gigapixel whole slide images with HistoGPT
Tran, M., Schmidle, P., Guo, R. R., Wagner, S. J., Koch, V., Lupperger, V., Novotny, B., Murphree, D. H., Hardway, H. D., D’Amato, M., et al. (2025) · Nature Communications · doi:10.1038/s41467-025-60014-x · Cited by 36 · see citing works
“Wetypicallyprocess 1000 to 10,000 patches per slide, although this number can vary widely as some slides may contain only a few hundred spots while others may contain tens of thousands.”
Zhu, M., Tian, Y., Chen, H., Zhou, C., Guo, Q., Liu, Y., Yang, M., & Shen, C. (2025) · 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52734.2025.00349 · Cited by 5 · see citing works
BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games
Paglieri, D., Cupiał, B., Coward, S., Piterbarg, U., Wolczyk, M., Khan, A., Pignatelli, E., Kuciński, Ł., Pinto, L., Fergus, R., et al. (2024) · arXiv · doi:10.48550/arxiv.2411.13543 · Cited by 2 · see citing works
“…with our findings, including biases toward natural image-text pairs, optimization for image description rather than action-oriented reasoning, and challenges with out-of-distribution inputs (Tan et al., 2024; Tong et al., 2024; Rah-manzadehgervi et al., 2024; Zang et al., 2024; Guan et al., 2023).”
Does Spatial Cognition Emerge in Frontier Models?
Ramakrishnan, S. K., Wijmans, E., Kraehenbuehl, P., & Koltun, V. (2024) · arXiv · doi:10.48550/arxiv.2410.06468
- Authors
- Tong, S., Liu, Z., Zhai, Y., Ma, Y., LeCun, Y., & Xie, S.
- Affiliations
- New York University (US)
- FAIR, Meta
- University of California, Berkeley (US)
- Venue
- 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
- Publisher
- IEEE
- Published
- 2024-06-16
- Pages
- 9568-9578
- ISSN
- 1063-6919 (ISSN-L)
- OpenAlex
- W4402754023
- Semantic Scholar
- ca00f4056f9039d3c1a4c3a113f5ee0527149b66
- Cited by
- 189
- References
- 48
- Citations by year
- 2026: 76 · 2025: 80 · 2024: 33
- FWCI
- 39.15
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings (publishedVersion)
- arXiv (Cornell University) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Keywords
- multimodal large language models, CLIP, visual grounding, GPT-4V, hallucinations, visual representation learning, vision-language models
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
