{"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/decoupled-kullback-leibler-divergence-loss","title":"Decoupled Kullback-Leibler Divergence Loss","arxiv_id":"2305.13948","date":"2023-05-23","proceeding":null,"authors":["Jiequan Cui","Zhuotao Tian","Zhisheng Zhong","Xiaojuan Qi","Bei Yu","Hanwang Zhang"],"abstract":"In this paper, we delve deeper into the Kullback-Leibler (KL) Divergence loss and mathematically prove that it is equivalent to the Decoupled Kullback-Leibler (DKL) Divergence loss that consists of 1) a weighted Mean Square Error (wMSE) loss and 2) a Cross-Entropy loss incorporating soft labels. Thanks to the decomposed formulation of DKL loss, we have identified two areas for improvement. Firstly, we address the limitation of KL/DKL in scenarios like knowledge distillation by breaking its asymmetric optimization property. This modification ensures that the $\\mathbf{w}$MSE component is always effective during training, providing extra constructive cues. Secondly, we introduce class-wise global information into KL/DKL to mitigate bias from individual samples. With these two enhancements, we derive the Improved Kullback-Leibler (IKL) Divergence loss and evaluate its effectiveness by conducting experiments on CIFAR-10/100 and ImageNet datasets, focusing on adversarial training, and knowledge distillation tasks. The proposed approach achieves new state-of-the-art adversarial robustness on the public leaderboard -- RobustBench and competitive performance on knowledge distillation, demonstrating the substantial practical merits. Our code is available at https://github.com/jiequancui/DKL.","url_abs":"https://arxiv.org/abs/2305.13948v3","url_pdf":"https://arxiv.org/pdf/2305.13948v3.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":"decoupled-kullback-leibler-divergence-loss","repo_url":"https://github.com/jiequancui/DKL","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"decoupled-kullback-leibler-divergence-loss","repo_url":"https://github.com/FPNAS/LBGAT","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"decoupled-kullback-leibler-divergence-loss","repo_url":"https://github.com/dvlab-research/LBGAT","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"decoupled-kullback-leibler-divergence-loss","repo_url":"https://github.com/jiequancui/LBGAT","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"adversarial-defense","task_name":"Adversarial Defense"},{"task_slug":"adversarial-robustness","task_name":"Adversarial Robustness"},{"task_slug":"knowledge-distillation","task_name":"Knowledge Distillation"}],"methods":[{"method_slug":"dkl","method_name":"DKL"},{"method_slug":"knowledge-distillation","method_name":"Knowledge Distillation"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2305.13948","atlas_url":"https://app.syntology.ai/?focus=2305.13948","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2305.13948"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+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/jiequancui/DKL","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/jiequancui/LBGAT","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/FPNAS/LBGAT","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/dvlab-research/LBGAT","reach":null}],"summary":{"ran":1,"ran_draft_wrong":6,"unverified":4},"by_repo_kind":{"official":{"samples":4,"ran":4,"repositories":1},"listed":{"samples":7,"ran":3,"repositories":3}},"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":4,"samples":[{"code_sha256_prefix":"b32e10adc9bf6201","entry":"IKL_KD","repo":"jiequancui/DKL","repo_kind":"official","path":"DKLv1/KD-dkl/mdistiller/distillers/IKL_KD.py","file_url":"https://github.com/jiequancui/DKL/blob/HEAD/DKLv1/KD-dkl/mdistiller/distillers/IKL_KD.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"b32e10adc9bf6201"}},{"code_sha256_prefix":"59f6b4ce1de254e2","entry":"cross_entropy","repo":"jiequancui/dkl","repo_kind":"official","path":"DKLv1/Adv-training-dkl/train_dkl_cifar10.py","file_url":"https://github.com/jiequancui/dkl/blob/HEAD/DKLv1/Adv-training-dkl/train_dkl_cifar10.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"59f6b4ce1de254e2"}},{"code_sha256_prefix":"eb1d0d5bb0f0c627","entry":"dkl_loss","repo":"jiequancui/dkl","repo_kind":"official","path":"DKLv1/Adv-training-dkl/train_dkl_cifar10.py","file_url":"https://github.com/jiequancui/dkl/blob/HEAD/DKLv1/Adv-training-dkl/train_dkl_cifar10.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"eb1d0d5bb0f0c627"}},{"code_sha256_prefix":"6ae3a3d2a8a1acc4","entry":"dkl_loss","repo":"jiequancui/dkl","repo_kind":"official","path":"DKLv1/KD-dkl/mdistiller/distillers/IKL_KD.py","file_url":"https://github.com/jiequancui/dkl/blob/HEAD/DKLv1/KD-dkl/mdistiller/distillers/IKL_KD.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"6ae3a3d2a8a1acc4"}},{"code_sha256_prefix":"bbffec6aadbbfd95","entry":"l2_norm","repo":"FPNAS/LBGAT","repo_kind":"listed","path":"lbgat.py","file_url":"https://github.com/FPNAS/LBGAT/blob/HEAD/lbgat.py","link_basis":"plan_row","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"bbffec6aadbbfd95"}},{"code_sha256_prefix":"342a07cb13a5e0bc","entry":"lbgat_loss","repo":"dvlab-research/LBGAT","repo_kind":"listed","path":"lbgat.py","file_url":"https://github.com/dvlab-research/LBGAT/blob/HEAD/lbgat.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"342a07cb13a5e0bc"}},{"code_sha256_prefix":"724c88e3f81d2123","entry":"squared_l2_norm","repo":"FPNAS/LBGAT","repo_kind":"listed","path":"lbgat.py","file_url":"https://github.com/FPNAS/LBGAT/blob/HEAD/lbgat.py","link_basis":"plan_row","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"724c88e3f81d2123"}},{"code_sha256_prefix":"87fe041ab6db516b","entry":"ResNet152_cifar","repo":"jiequancui/LBGAT","repo_kind":"listed","path":"models/resnetcifar.py","file_url":"https://github.com/jiequancui/LBGAT/blob/HEAD/models/resnetcifar.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":"87fe041ab6db516b"}},{"code_sha256_prefix":"60d3d2bb8b6c30e2","entry":"ResNet18_cifar","repo":"jiequancui/LBGAT","repo_kind":"listed","path":"models/resnetcifar.py","file_url":"https://github.com/jiequancui/LBGAT/blob/HEAD/models/resnetcifar.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":"60d3d2bb8b6c30e2"}},{"code_sha256_prefix":"4386c6d158344286","entry":"ResNet34_cifar","repo":"jiequancui/LBGAT","repo_kind":"listed","path":"models/resnetcifar.py","file_url":"https://github.com/jiequancui/LBGAT/blob/HEAD/models/resnetcifar.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":"4386c6d158344286"}},{"code_sha256_prefix":"2533491b22247ed9","entry":"lbgat_loss","repo":"jiequancui/LBGAT","repo_kind":"listed","path":"lbgat.py","file_url":"https://github.com/jiequancui/LBGAT/blob/HEAD/lbgat.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":"2533491b22247ed9"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}