{"ID":6537715,"CreatedAt":"2026-07-14T02:54:43.516908796Z","UpdatedAt":"2026-07-15T03:28:55.185153975Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.11163","arxiv_id":"2607.11163","title":"Unified Gradient Projection: Language-Balanced Continual Learning for Multilingual Low-Resource ASR","abstract":"Large-scale pretrained ASR models such as Whisper exhibit strong multilingual capabilities. However, fine-tuning on low-resource languages often causes catastrophic forgetting. Although continual learning mitigates this issue, existing methods struggle to regulate cross-task interference in multilingual settings, where dominant languages bias optimization. We propose Unified Gradient Projection (UGP), which constrains parameter updates using reference gradients from language-balanced replay in a unified projection space. By equalizing per-language contributions in the projection, UGP reduces dominant-language bias and improves cross-lingual stability. We further show that combining gradient-level projection with data-level replay yields complementary gains in stability and plasticity. Across diverse low-resource language groups and model scales, UGP enables effective adaptation while substantially mitigating forgetting. On Whisper-large-v3, it achieves near-zero average forgetting.","short_abstract":"Large-scale pretrained ASR models such as Whisper exhibit strong multilingual capabilities. However, fine-tuning on low-resource languages often causes catastrophic forgetting. Although continual learning mitigates this issue, existing methods struggle to regulate cross-task interference in multilingual settings, where...","url_abs":"https://arxiv.org/abs/2607.11163","url_pdf":"https://arxiv.org/pdf/2607.11163v1","authors":"[\"Ziang Ren\",\"Guodong Lin\",\"Yuchen Ai\",\"Kaize Tan\",\"Wei-Qiang Zhang\"]","published":"2026-07-13T06:57:02Z","proceeding":"cs.CL","tasks":"[\"cs.CL\",\"cs.SD\",\"eess.AS\"]","methods":"[]","has_code":false}