{"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/arxiv-2602-06883","title":"Vision Transformer Finetuning Benefits from Non-Smooth Components","arxiv_id":"2602.06883","date":"2026-02-06","proceeding":"ICML","authors":["Ambroise Odonnat","Laetitia Chapel","Romain Tavenard","Ievgen Redko"],"abstract":"The smoothness of the transformer architecture has been extensively studied in the context of generalization, training stability, and adversarial robustness. However, its role in transfer learning remains poorly understood. In this paper, we analyze the ability of vision transformer components to adapt their outputs to changes in inputs, or, in other words, their \\emph{plasticity}. Defined as an average rate of change, it captures the sensitivity to input perturbation; in particular, a high plasticity implies a low smoothness. Our theoretical analysis and extensive experiments -- over $1,000$ finetuning runs on large-scale vision transformers -- showcase that this perspective provides principled guidance in choosing the components to prioritize during adaptation. A key takeaway for practitioners is that the high plasticity of the attention modules and feedforward layers consistently leads to better finetuning performance. Our findings depart from the prevailing assumption that smoothness is desirable, offering a novel perspective on transformers' functional properties. The code is available at https://github.com/ambroiseodt/vit-plasticity.","url_abs":"https://arxiv.org/abs/2602.06883","url_pdf":"https://arxiv.org/pdf/2602.06883","source":{"archive":null,"snapshot":"2025-07-28","note":"not in the Papers with Code archive (frozen at the snapshot)","row_kind":"graph","title_abstract_authors_date":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)"},"code_links":[],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=2602.06883","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2602.06883"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"mentioned_in_github":null,"is_official":null,"provenance":"deterministic:regex_extraction","mentioned_in_paper":null,"url":"https://github.com/ambroiseodt/vit-plasticity","reach":null}],"summary":{"unverified":10},"by_repo_kind":{"found_in_text":{"samples":10,"ran":0,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"0c8eb7277387806b","entry":"build_optimizer","repo":"ambroiseodt/vit-plasticity","repo_kind":"found_in_text","path":"src/vitef/optim.py","file_url":"https://github.com/ambroiseodt/vit-plasticity/blob/HEAD/src/vitef/optim.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"0c8eb7277387806b"}},{"code_sha256_prefix":"c9981055b9019584","entry":"build_with_type_check","repo":"ambroiseodt/vit-plasticity","repo_kind":"found_in_text","path":"src/vitef/utils.py","file_url":"https://github.com/ambroiseodt/vit-plasticity/blob/HEAD/src/vitef/utils.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"c9981055b9019584"}},{"code_sha256_prefix":"2b47253f170c2a89","entry":"get_jsonl_keys","repo":"ambroiseodt/vit-plasticity","repo_kind":"found_in_text","path":"src/vitef/utils.py","file_url":"https://github.com/ambroiseodt/vit-plasticity/blob/HEAD/src/vitef/utils.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"2b47253f170c2a89"}},{"code_sha256_prefix":"d151a2ea39059f9c","entry":"get_pca_basis","repo":"ambroiseodt/vit-plasticity","repo_kind":"found_in_text","path":"apps/plots/loss_landscape.py","file_url":"https://github.com/ambroiseodt/vit-plasticity/blob/HEAD/apps/plots/loss_landscape.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"d151a2ea39059f9c"}},{"code_sha256_prefix":"c4ee7ffac8194eb3","entry":"get_raw_model","repo":"ambroiseodt/vit-plasticity","repo_kind":"found_in_text","path":"src/vitef/distributed.py","file_url":"https://github.com/ambroiseodt/vit-plasticity/blob/HEAD/src/vitef/distributed.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"c4ee7ffac8194eb3"}},{"code_sha256_prefix":"e0deed7ee32dc0c5","entry":"interpolate_frames","repo":"ambroiseodt/vit-plasticity","repo_kind":"found_in_text","path":"apps/plots/loss_landscape.py","file_url":"https://github.com/ambroiseodt/vit-plasticity/blob/HEAD/apps/plots/loss_landscape.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"e0deed7ee32dc0c5"}},{"code_sha256_prefix":"d854539c5e14110e","entry":"load_jsonl_to_numpy","repo":"ambroiseodt/vit-plasticity","repo_kind":"found_in_text","path":"src/vitef/utils.py","file_url":"https://github.com/ambroiseodt/vit-plasticity/blob/HEAD/src/vitef/utils.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"d854539c5e14110e"}},{"code_sha256_prefix":"4c20b73b071752f2","entry":"lr_constant","repo":"ambroiseodt/vit-plasticity","repo_kind":"found_in_text","path":"src/vitef/optim.py","file_url":"https://github.com/ambroiseodt/vit-plasticity/blob/HEAD/src/vitef/optim.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"4c20b73b071752f2"}},{"code_sha256_prefix":"712ef77b64ba510b","entry":"lr_linear","repo":"ambroiseodt/vit-plasticity","repo_kind":"found_in_text","path":"src/vitef/optim.py","file_url":"https://github.com/ambroiseodt/vit-plasticity/blob/HEAD/src/vitef/optim.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"712ef77b64ba510b"}},{"code_sha256_prefix":"6a3bc5a7f790da13","entry":"set_seed","repo":"ambroiseodt/vit-plasticity","repo_kind":"found_in_text","path":"src/vitef/config.py","file_url":"https://github.com/ambroiseodt/vit-plasticity/blob/HEAD/src/vitef/config.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"6a3bc5a7f790da13"}}]},"arxiv_metadata":{"licence":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)","fields":["title","abstract","authors","date"],"primary_category":"cs.LG","source":"arxiv_2026.jsonl"},"syntology_extracted_results":null}