{"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/multi-criteria-token-fusion-with-one-step","title":"Multi-criteria Token Fusion with One-step-ahead Attention for Efficient Vision Transformers","arxiv_id":"2403.10030","date":"2024-03-15","proceeding":"CVPR 2024 1","authors":["Sanghyeok Lee","Joonmyung Choi","Hyunwoo J. Kim"],"abstract":"Vision Transformer (ViT) has emerged as a prominent backbone for computer vision. For more efficient ViTs, recent works lessen the quadratic cost of the self-attention layer by pruning or fusing the redundant tokens. However, these works faced the speed-accuracy trade-off caused by the loss of information. Here, we argue that token fusion needs to consider diverse relations between tokens to minimize information loss. In this paper, we propose a Multi-criteria Token Fusion (MCTF), that gradually fuses the tokens based on multi-criteria (e.g., similarity, informativeness, and size of fused tokens). Further, we utilize the one-step-ahead attention, which is the improved approach to capture the informativeness of the tokens. By training the model equipped with MCTF using a token reduction consistency, we achieve the best speed-accuracy trade-off in the image classification (ImageNet1K). Experimental results prove that MCTF consistently surpasses the previous reduction methods with and without training. Specifically, DeiT-T and DeiT-S with MCTF reduce FLOPs by about 44% while improving the performance (+0.5%, and +0.3%) over the base model, respectively. We also demonstrate the applicability of MCTF in various Vision Transformers (e.g., T2T-ViT, LV-ViT), achieving at least 31% speedup without performance degradation. Code is available at https://github.com/mlvlab/MCTF.","url_abs":"https://arxiv.org/abs/2403.10030v3","url_pdf":"https://arxiv.org/pdf/2403.10030v3.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":[{"paper_slug":"multi-criteria-token-fusion-with-one-step","repo_url":"https://github.com/mlvlab/mctf","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"computational-efficiency","task_name":"Computational Efficiency"},{"task_slug":"efficient-vits","task_name":"Efficient ViTs"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[{"method_slug":"deit","method_name":"DeiT"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"lv-vit","method_name":"LV-ViT"},{"method_slug":"pruning","method_name":"Pruning"},{"method_slug":"t2t-vit","method_name":"T2T-ViT"},{"method_slug":"vision-transformer","method_name":"Vision Transformer"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/efficient-vits-on-imagenet-1k-with-lv-vit-s","task":"Efficient ViTs","dataset":"ImageNet-1K (With LV-ViT-S)","model":"MCTF ($r=8$)","rank_in_archive_order":1,"of":19,"metrics":{"GFLOPs":"4.9","Top 1 Accuracy":"83.5"},"uses_additional_data":false},{"leaderboard":"/sota/efficient-vits-on-imagenet-1k-with-lv-vit-s","task":"Efficient ViTs","dataset":"ImageNet-1K (With LV-ViT-S)","model":"MCTF ($r=12$)","rank_in_archive_order":2,"of":19,"metrics":{"GFLOPs":"4.2","Top 1 Accuracy":"83.4"},"uses_additional_data":false},{"leaderboard":"/sota/efficient-vits-on-imagenet-1k-with-lv-vit-s","task":"Efficient ViTs","dataset":"ImageNet-1K (With LV-ViT-S)","model":"MCTF ($r=16$)","rank_in_archive_order":19,"of":19,"metrics":{"GFLOPs":"3.6","Top 1 Accuracy":"82.3"},"uses_additional_data":false},{"leaderboard":"/sota/efficient-vits-on-imagenet-1k-with-deit-s","task":"Efficient ViTs","dataset":"ImageNet-1K (with DeiT-S)","model":"MCTF ($r=16$)","rank_in_archive_order":1,"of":41,"metrics":{"GFLOPs":"2.6","Top 1 Accuracy":"80.1"},"uses_additional_data":false},{"leaderboard":"/sota/efficient-vits-on-imagenet-1k-with-deit-s","task":"Efficient ViTs","dataset":"ImageNet-1K (with DeiT-S)","model":"MCTF ($r=18$)","rank_in_archive_order":3,"of":41,"metrics":{"GFLOPs":"2.4","Top 1 Accuracy":"79.9"},"uses_additional_data":false},{"leaderboard":"/sota/efficient-vits-on-imagenet-1k-with-deit-s","task":"Efficient ViTs","dataset":"ImageNet-1K (with DeiT-S)","model":"MCTF ($r=20$)","rank_in_archive_order":19,"of":41,"metrics":{"GFLOPs":"2.2","Top 1 Accuracy":"79.5"},"uses_additional_data":false},{"leaderboard":"/sota/efficient-vits-on-imagenet-1k-with-deit-t","task":"Efficient ViTs","dataset":"ImageNet-1K (with DeiT-T)","model":"MCTF ($r=8$)","rank_in_archive_order":2,"of":22,"metrics":{"GFLOPs":"1.0","Top 1 Accuracy":"72.9"},"uses_additional_data":false},{"leaderboard":"/sota/efficient-vits-on-imagenet-1k-with-deit-t","task":"Efficient ViTs","dataset":"ImageNet-1K (with DeiT-T)","model":"MCTF ($r=16$)","rank_in_archive_order":3,"of":22,"metrics":{"GFLOPs":"0.7","Top 1 Accuracy":"72.7"},"uses_additional_data":false},{"leaderboard":"/sota/efficient-vits-on-imagenet-1k-with-deit-t","task":"Efficient ViTs","dataset":"ImageNet-1K (with DeiT-T)","model":"MCTF ($r=20$)","rank_in_archive_order":16,"of":22,"metrics":{"GFLOPs":"0.6","Top 1 Accuracy":"71.4"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2403.10030","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2403.10030"}},"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":[{"provenance":"deterministic:regex_extraction","url":"https://github.com/mlvlab/MCTF","reach":null}],"summary":{"ran":3,"unverified":2},"by_repo_kind":{"official":{"samples":5,"ran":3,"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":"00e9d5d63d4079de","entry":"do_nothing","repo":"mlvlab/MCTF","repo_kind":"official","path":"utils/mctf.py","file_url":"https://github.com/mlvlab/MCTF/blob/HEAD/utils/mctf.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"00e9d5d63d4079de"}},{"code_sha256_prefix":"8a67c271b170dd65","entry":"merge_source","repo":"mlvlab/MCTF","repo_kind":"official","path":"utils/mctf.py","file_url":"https://github.com/mlvlab/MCTF/blob/HEAD/utils/mctf.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"8a67c271b170dd65"}},{"code_sha256_prefix":"35fd12a5c86d3087","entry":"merge_wavg","repo":"mlvlab/MCTF","repo_kind":"official","path":"utils/mctf.py","file_url":"https://github.com/mlvlab/MCTF/blob/HEAD/utils/mctf.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"35fd12a5c86d3087"}},{"code_sha256_prefix":"58a3d8aacef250bc","entry":"bipartite_soft_matching","repo":"mlvlab/MCTF","repo_kind":"official","path":"utils/mctf.py","file_url":"https://github.com/mlvlab/MCTF/blob/HEAD/utils/mctf.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":"58a3d8aacef250bc"}},{"code_sha256_prefix":"92d18f3cf6ae00a5","entry":"mctf","repo":"mlvlab/MCTF","repo_kind":"official","path":"utils/mctf.py","file_url":"https://github.com/mlvlab/MCTF/blob/HEAD/utils/mctf.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":"92d18f3cf6ae00a5"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}