{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,18]],"date-time":"2026-03-18T03:15:07Z","timestamp":1773803707528,"version":"3.50.1"},"reference-count":0,"publisher":"Association for the Advancement of Artificial Intelligence (AAAI)","issue":"30","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["AAAI"],"abstract":"<jats:p>Singular Value Decomposition (SVD) has recently gained traction as an effective compression technique for large language models (LLMs), with many studies reporting 20-80% parameter reduction at minimal accuracy cost. However, despite reducing weight memory, existing SVD-based approaches still rely on standard dense CUDA kernels during inference, which incur substantial-and ultimately unnecessary-activation memory overhead. Our analysis reveals that this kernel-induced cost, which grows with sequence length and hidden size, in worst case prevents any real reduction in peak inference memory, limiting the practical impact of SVD compression for on-device deployment. To address this bottleneck, we propose FlashSVD, an end-to-end, rank-aware streaming inference framework for SVD-compressed LLMs. FlashSVD integrates seamlessly with any SVD-based model and directly fuses low-rank projection kernels into self-attention and feed-forward pipelines. This design avoids materializing large activation buffers by streaming small tiles of truncated factors through on-chip SRAM, performing on-the-fly multiplication and reduction, and immediately evicting results\u2013thus preserving high GPU occupancy without introducing latency. On standard benchmarks (e.g., BERT-Base), FlashSVD reduces peak activation memory by up to 70.2% and transient memory by 75%, with zero accuracy loss against low-rank baselines, enabling truly memory-efficient deployment of low-rank LLMs.<\/jats:p>","DOI":"10.1609\/aaai.v40i30.39720","type":"journal-article","created":{"date-parts":[[2026,3,18]],"date-time":"2026-03-18T02:03:33Z","timestamp":1773799413000},"page":"25278-25285","source":"Crossref","is-referenced-by-count":0,"title":["FlashSVD: Memory-Efficient Inference with Streaming for Low-Rank Models"],"prefix":"10.1609","volume":"40","author":[{"given":"Zishan","family":"Shao","sequence":"first","affiliation":[]},{"given":"Yixiao","family":"Wang","sequence":"additional","affiliation":[]},{"given":"Qinsi","family":"Wang","sequence":"additional","affiliation":[]},{"given":"Ting","family":"Jiang","sequence":"additional","affiliation":[]},{"given":"Zhixu","family":"Du","sequence":"additional","affiliation":[]},{"given":"Hancheng","family":"Ye","sequence":"additional","affiliation":[]},{"given":"Danyang","family":"Zhuo","sequence":"additional","affiliation":[]},{"given":"Yiran","family":"Chen","sequence":"additional","affiliation":[]},{"given":"Hai","family":"\u00a8Helen\u00a8 Li","sequence":"additional","affiliation":[]}],"member":"9382","published-online":{"date-parts":[[2026,3,14]]},"container-title":["Proceedings of the AAAI Conference on Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/download\/39720\/43681","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/download\/39720\/43681","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,18]],"date-time":"2026-03-18T02:03:33Z","timestamp":1773799413000},"score":1,"resource":{"primary":{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/view\/39720"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,3,14]]},"references-count":0,"journal-issue":{"issue":"30","published-online":{"date-parts":[[2026,3,17]]}},"URL":"https:\/\/doi.org\/10.1609\/aaai.v40i30.39720","relation":{},"ISSN":["2374-3468","2159-5399"],"issn-type":[{"value":"2374-3468","type":"electronic"},{"value":"2159-5399","type":"print"}],"subject":[],"published":{"date-parts":[[2026,3,14]]}}}