{"paper":{"title":"FedHPro: Federated Hyper-Prototype Learning via Gradient Matching","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Hyper-prototypes aligned by gradient matching from client samples reduce semantic drift in federated prototype learning.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Di Wu, Guansong Pang, Haoran Li, Huan Wang, Jun Shen, Jun Yan, Ousman Manjang, Yanlong Zhai, Zhenyu Yang","submitted_at":"2026-05-13T13:01:17Z","abstract_excerpt":"Federated Learning (FL) enables collaborative training of distributed clients while protecting privacy. To enhance generalization capability in FL, prototype-based FL is in the spotlight, since shared global prototypes offer semantic anchors for aligning client-specific local prototypes. However, existing methods update global prototypes at the prototype-level via averaging local prototypes or refining global anchors, which often leads to semantic drift across clients and subsequently yields a misaligned global signal. To alleviate this issue, we introduce hyper-prototypes, defined by a set of"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"hyper-prototypes produce a more semantically consistent global signal, and FedHPro achieves state-of-the-art performance on several benchmark datasets under diverse heterogeneous scenarios.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That matching gradients computed on real client samples will align hyper-prototypes more reliably than averaging local prototypes, without introducing new privacy leakage or optimization instability.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"FedHPro introduces gradient-matched hyper-prototypes plus mutual-contrastive learning to produce semantically consistent global signals and reach state-of-the-art accuracy on heterogeneous image benchmarks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Hyper-prototypes aligned by gradient matching from client samples reduce semantic drift in federated prototype learning.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"73c7775217d26fde24604929157ff595b8e06f2484f2c39e73f1ef52fe8417fe"},"source":{"id":"2605.13475","kind":"arxiv","version":1},"verdict":{"id":"caa6e760-881e-4e95-a58d-e6bf01719f99","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T20:21:59.453163Z","strongest_claim":"hyper-prototypes produce a more semantically consistent global signal, and FedHPro achieves state-of-the-art performance on several benchmark datasets under diverse heterogeneous scenarios.","one_line_summary":"FedHPro introduces gradient-matched hyper-prototypes plus mutual-contrastive learning to produce semantically consistent global signals and reach state-of-the-art accuracy on heterogeneous image benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That matching gradients computed on real client samples will align hyper-prototypes more reliably than averaging local prototypes, without introducing new privacy leakage or optimization instability.","pith_extraction_headline":"Hyper-prototypes aligned by gradient matching from client samples reduce semantic drift in federated prototype learning."},"references":{"count":14,"sample":[{"doi":"","year":2066,"title":"Geodesic flow kernel for unsupervised domain adaptation","work_id":"30078da4-5702-4a95-b330-9b6301ace13e","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2062,"title":"Learning support and trivial prototypes for interpretable image classification","work_id":"3f34da05-f0b7-4392-b5ea-30fb79842983","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"12 FedHPro: Federated Hyper-Prototype Learning via Gradient Matching A. Algorithm Pseudo-code Flow In this section, we describe the pseudo-code of our FedHPro in Algorithm 1: 1)Server-Side: we optimiz","work_id":"cb75084a-2f4e-4297-b7a1-619b6b206eaa","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"for the TinyImageNet dataset. The label skew heterogeneity level of clients is controlled by the standard deviation α of the Dirichlet distribution, and the quantity skew heterogeneity level is contro","work_id":"08a17d60-2f9c-40f2-84a3-f6a4f704bbef","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2019,"title":"is used for skin lesion classification and contains 8,912 training samples and 1,103 testing samples with 7 categories, and each sample’s size is scaled to224∗224 . Then, based on (Kaidi et al., 2019)","work_id":"240b0b88-c1f8-430d-9032-d066a7b7ebd2","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":14,"snapshot_sha256":"a2362fba15b17627156b60fc7ae93c8e4c91c4648703dc75cb5d55c004a6c21b","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"543ef28dff658b3cec64dfa8e2d373a2576919cce2fc5e0ed023d3fed296c362"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}