What matters when building vision-language models?
Open accessThe DOI matches its Crossref record; 9 of the 9 references it cites were found in a registry or an index.
Abstract
The growing interest in vision-language models (VLMs) has been driven by improvements in large language models and vision transformers. Despite the abundance of literature on this subject, we observe that critical decisions regarding the design of VLMs are often not justified. We argue that these unsupported decisions impede progress in the field by making it difficult to identify which choices improve model performance. To address this issue, we conduct extensive experiments around pre-trained models, architecture choice, data, and training methods. Our consolidation of findings includes the development of Idefics2, an efficient foundational VLM of 8 billion parameters. Idefics2 achieves state-of-the-art performance within its size category across various multimodal benchmarks, and is often on par with models four times its size. We release the model (base, instructed, and chat) along with the datasets created for its training.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
The publisher hasn’t deposited this paper’s reference list.
Cited by
Items 601–660 of 660
A Spatial–Channel Aware Adaptive Connector for Multimodal Multi-Layer Feature Fusion
Zhang, H., Zhao, C., Wang, F., Wu, X., Ma, K., Zhu, H., Liang, S., & Zhang, P. (2026) · International Conference on Computer Supported Cooperative Work in Design (CSCWD) · doi:10.1109/cscwd68734.2026.11582221
Ojuri, S., Chahal, A., Lane, N., Rahimian, F. P., & Stefano, A. D. (2026) · Lecture notes in civil engineering · doi:10.1007/978-981-95-8872-5_33
RADIO1D: Elastic Representations for Condensed Vision ModelingPreprint
Heinrich, G., Ranzinger, M., McCarthy, C., Bagrov, N., Khvedchenya, E., Catanzaro, B., Kautz, J., Tao, A., & Molchanov, P. A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.03624
Injecting reasoning into vision-language models via weight-decomposed merging
Zou, Y., Wang, L. Y., Zhao, Q., Wei, Y., Li, Y., & Du, B. (2026) · Neural Networks · doi:10.1016/j.neunet.2026.109364
Mizutani, K., Yashiro, R., & Tokuyama, K. (2026) · Journal of Bioscience and Bioengineering · doi:10.1016/j.jbiosc.2026.06.003
de Rodrigo, I., López-López, Á. J., & Boal, J. (2026) · Pattern Recognition · doi:10.1016/j.patcog.2026.114448
Massively Multilingual Vision-Language Understanding
Geigle, G. (2026) · doi:10.25972/opus-44667
Le, X. T., Vo, Q. N. T., Tran, T. H. M., Nguyen, T. T., & Quan, T. T. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-2885-0_8
Design LoRA: Interactive fine-tuning of text-to-image models with designer-in-the-loop feedback
Lee, S. W., Yun, Y., & Hyun, K. H. (2026) · Advanced Engineering Informatics · doi:10.1016/j.aei.2026.105053
Towards joint scaling laws with optimal batch size schedulesPreprint
Li, J., Bu, Z., & Xu, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.27731
Lu, E. (2026) · Journal of Knowledge Learning and Science Technology · doi:10.60087/jklst.vol5.n2.005
Mukhametsharip, Z., Demberg, V., & Suresh, V. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.09772
Dynamic Distribution-Aware Uncertainty Tracking in Vision-Language Representation Learning
Zhou, A., Jiang, Z., Cheng, Z., Wang, C., Yang, S., Chen, H., & Gu, Q. (2026) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3767308.3836486
When Is a Task Vector Enough? An Empirical Theory of Implicit Multimodal ICLPreprint
Li, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.13385
Srinivasan, A., Jena, P., Topale, A., Venna, V. K., & Sarvadevabhatla, R. K. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-36033-5_37
Divya, E., Kolluri, J., & Siripuri, K. (2026) · International Conference on Smart Systems and Inventive Technology (ICSSIT) · doi:10.1109/icssit69151.2026.11656493
X-MULTI: VLM-based Imaging Factor Disentanglement for Factor-Aware Image SynthesisPreprint
Godavarthy, S., Neuwirth-Trapp, M., Faasch, T.-F., Bieshaar, M., Moeller, M., Laerhoven, K. V., & Paudel, D. P. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.24563
Huang, J., & Xiao, L. (2026) · International Journal of Pattern Recognition and Artificial Intelligence · doi:10.1142/s0218001426400392
Activation Outliers Matter: Robust Recovery for Quantized Multimodal LLMsPreprint
Rahman, T., Taghian, M., Peng, Y., Ma, Z., Raju, A. D., Wang, Y., Huang, X., Mak, H. Y., Golestan, S., Le, H., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.26581
Xiao, L., Wang, X., Ong, K., Wei, J., Dong, C., Mao, R., Cambria, E., & Mengaldo, G. (2026) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3767308.3836120
Song, Y., Zhang, S., Chai, Y., Wang, H., She, Y., Li, J., Gao, L., & Lan, M. (2026) · International Conference on Computer Information and Big Data Applications (CIBDA) · doi:10.1145/3813822.3814176
Das, A., Biswas, S., Chattopadhyay, S., Lodh, A., Pal, A., Banerjee, P., Singh, N., Jawahar, C. V., Lladós, J., & Karatzas, D. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37676-3_40
Fletcher, J. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7419558
Venna, V. K., Bandarupalli, S., Srinivasan, A., Raghuveer, R., Gunda, S. M., & Sarvadevabhatla, R. K. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37016-7_15
(How) Do MLLMs Report Bistable Images Like Humans?
Takatsuki, R., Doi, T., Watahiki, A., Seth, A. K., & Yanaka, H. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.1744
CMJU: cross-modal joint unlearning for balanced forgetting in multimodal large language models
Yang, D., Xiong, B., & Yang, X. (2026) · Visual Intelligence · doi:10.1007/s44267-026-00130-4
Linking Vision and Meaning: Text-Image Correspondence in Cultural Heritage Documentation
Chen, C., Tomko, M., & Khoshelham, K. (2026) · Journal on Computing and Cultural Heritage · doi:10.1145/3848029
Pandey, A., Mohan, A. K., & Sankar, P. S. B. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37314-4_36
Combe, C. G. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.20068
Fingerprinting Multimodal Large Language Models
Huang, C., Tong, M., & Chen, K.-J. (2026) · Proceedings of the 34th ACM International Conference on Multimedia · doi:10.1145/3767308.3834994
Wei, X., Zhou, D., Guo, Z., & Lu, Y. (2026) · Computers & Industrial Engineering · doi:10.1016/j.cie.2026.112392
Head-level lesion-symptom mapping of picture naming in vision-language modelsPreprint
Nemati, S., Newman‐Norlund, R., Ahmadi, S., Guan, X., Warren, K., Yang, Y., Nelakuditi, S., Rorden, C., Bonilha, L., & Fridriksson, J. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7513276
Liu, X., Zheng, H., Pu, N., Li, W., Sebe, N., & Zhong, Z. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37461-5_17
Limaroon, K., Namwongsa, P., & Tarnpradab, S. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-38062-3_25
When Text Matters: Design Principles for Visual Token Pruning in Vision-Language ModelPreprint
Kang, M., Park, K., Cho, S., Kim, D., & Cho, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.34861
Adapting Vision-Language Models for Human-Readable XAI in Industrial Object DetectionPreprint
Sardari, S., Fernandes, F., Yelvande, S., Araya-Martinez, J. M., & Roitberg, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.31690
VISOR++ : Visual Input Based Steering for Large Vision Language Models
Balakrishnan, R., & Phute, M. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37086-0_27
Two Vectors Replace In-Context Demos: Structured Task Adaptation via EmbeddingsPreprint
Ding, X., Shi, N., & Zhang, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.07572
Selective Transfer of RL Updates for Visual ReasoningPreprint
Ji, S., Wan, H., Liu, M., & Zhang, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.08659
GeoPID: Decomposing and Steering Visual Information in Vision-Language ModelsPreprint
Kim, S., Zhang, Z., Zhang, X., Ling, J., & Shaw, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.08401
Loureiro, M., Dias, J., Melo, B., Sanchotene, N., Salles, D., Santini, R. M., & Ferreira, F. (2026) · International journal of communication · doi:10.65476/s92qta53
Zhang, S., Zeng, W., Zhang, Z., Tang, J., & Zhao, X. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.14542
Rajabi, M. S. (2026) · VTechWorks (Virginia Tech)
Kasoura, E. E., Damianos, D., Paraskevopoulos, G., & Katsouros, V. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.19932
Shift-Reliance Alignment: Ranking Vision-Language Models by the Cues They TrustPreprint
Mohammad, F., & Younis, M. (2026) · arXiv (Cornell University) · doi:10.13016/m2rxuo-ubfk
SL-Ground: Sentence-Level Grounding for Fine-Grained Open-Vocabulary Object DetectionPreprint
Li, Y.-S., Yang, C.-H., Jui, C.-Y., & Lu, P. J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.24814
FlashAttention-Compatible Modality-Mutual Attention with Virtual Reordering
Seo, J. W., & Jung, H. (2026) · Mathematics · doi:10.3390/math14213852
Watch and Learn: Teaching Visual In-Context Learning by Removing Textual ShortcutsPreprint
Shabtay, N., Yehudai, A., Dekel, A., Schwartz, E., & Giryes, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.30913
Zheng, X., Wang, X., & Luo, M. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.2680
Anchored Reasoning on Tables: Bridging the Verifiability Gap in Multimodal Reinforcement Learning
Jiang, Y., An, Y., Xie, F., Jiang, Z., & Zhang, H. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.2306
Wang, X., Zheng, X., Peng, P., & Luo, M. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.2476
RI-Prune: Relation and Information Aware Visual Token Pruning for Multi-Image Vision-Language Models
Lee, D., Lee, E., Jang, J., Lee, Y., & Kim, Y. (2026) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3799682.3841123
Multimodal In-Context Learning with Explicit Rule InductionPreprint
Li, Y., Jiang, X., Xue, Y., & Zhao, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.41754
See, Think, Use: Diagnosing Perception and Reasoning Failures in VLMsPreprint
Ding, J., Qi, Y., Guo, C., & Mai, P. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2611.01467
PixelBench: Can LLMs Understand and Generate Images Pixel by Pixel?Preprint
Zhang, H., Liu, F., Tian, J., Li, W., Duan, H., & Liu, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2611.02540
Roy, A., Shukla, P., Cheng, J., Ghosh, S., Craig, P., Balaji, A., Phatak, M. V., Bahrami, H., & Asadizanzani, N. (2026) · IMAPSource Proceedings · doi:10.4071/001c.173631
Towards Generalized No-Reference Image Quality Assessment Via Ascend Acceleration Learning
Zhang, X., Hong, W., Lin, D., He, X., Zou, R., Zhou, Z., Wang, F., Mai, J., Wang, W., Xia, Y., et al. (2026) · International Conference on Intelligent Computing and Robotics (ICICR) · doi:10.1109/icicr69256.2026.00237
When and How to use In-context Learning for modern LVLMs? An Extensive Investigation
Li, P., Zhao, Y., Yu, J., Xie, L., & Yin, E. (2026) · doi:10.18653/2026.findings-aacl.162
Bayesian Low-Rank Posteriors for Scalable Membership InferencePreprint
Nguyen, K., Ciná, A. E., Barsky, A., Reinares, A., Roli, F., & Karatzas, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2611.10455
Elhanashi, A., Zheng, Q., & Saponara, S. (2026) · Applied Sciences · doi:10.3390/app162211147
Metadata
- Authors
- Laurençon, H., Tronchon, L., Cord, M., & Sanh, V.
- ORCID
- Cord, Matthieu
- Venue
- Advances in Neural Information Processing Systems 37
- Published
- 2024
- Pages
- 87874-87907
- ISSN
- 1049-5258 (ISSN-L)
- OpenAlex
- W4415796627
- Cited by
- 66
- References
- 0
- Citations by year
- 2026: 39 · 2025: 23 · 2024: 4
- FWCI
- 13.67
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- cc-by
- Locations
- neural information processing systems (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 › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Keywords
- vision-language models, model design, multimodal representation learning
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
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.