{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/what-matters-when-building-vision-language","title":"What matters when building vision-language models?","arxiv_id":"2405.02246","date":"2024-05-03","proceeding":null,"authors":["Hugo Laurençon","Léo Tronchon","Matthieu Cord","Victor Sanh"],"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.","url_abs":"https://arxiv.org/abs/2405.02246v1","url_pdf":"https://arxiv.org/pdf/2405.02246v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[],"tasks":[{"task_slug":"1-image-2-2-stitching","task_name":"1 Image, 2*2 Stitching"},{"task_slug":"image-retrieval","task_name":"Image Retrieval"},{"task_slug":"long-context-understanding","task_name":"Long-Context Understanding"},{"task_slug":"mmr-total","task_name":"MMR total"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/long-context-understanding-on-mmneedle","task":"Long-Context Understanding","dataset":"MMNeedle","model":"IDEFICS2-8B","rank_in_archive_order":7,"of":12,"metrics":{"1 Image, 2*2 Stitching, Exact Accuracy":"18.9","1 Image, 4*4 Stitching, Exact Accuracy":"7.8","1 Image, 8*8 Stitching, Exact Accuracy":"0.9","10 Images, 1*1 Stitching, Exact Accuracy":"0","10 Images, 2*2 Stitching, Exact Accuracy":"0","10 Images, 4*4 Stitching, Exact Accuracy":"0","10 Images, 8*8 Stitching, Exact Accuracy":"0"},"uses_additional_data":false},{"leaderboard":"/sota/mmr-total-on-mrr-benchmark","task":"MMR total","dataset":"MRR-Benchmark","model":"Idefics-2-8B","rank_in_archive_order":10,"of":14,"metrics":{"Total Column Score":"256"},"uses_additional_data":true}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2405.02246","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}