{"ID":2879336,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2508.16703","arxiv_id":"2508.16703","title":"ShadowNPU: System and Algorithm Co-design for NPU-Centric On-Device LLM Inference","abstract":"On-device running Large Language Models (LLMs) is nowadays a critical enabler towards preserving user privacy. We observe that the attention operator falls back from the special-purpose NPU to the general-purpose CPU/GPU because of quantization sensitivity in state-of-the-art frameworks. This fallback results in a degraded user experience and increased complexity in system scheduling. To this end, this paper presents shadowAttn, a system-algorithm codesigned sparse attention module with minimal reliance on CPU/GPU by only sparsely calculating the attention on a tiny portion of tokens. The key idea is to hide the overhead of estimating the important tokens with a NPU-based pilot compute. Further, shadowAttn proposes insightful techniques such as NPU compute graph bucketing, head-wise NPU-CPU/GPU pipeline and per-head fine-grained sparsity ratio to achieve high accuracy and efficiency. shadowAttn delivers the best performance with highly limited CPU/GPU resource; it requires much less CPU/GPU resource to deliver on-par performance of SoTA frameworks.","short_abstract":"On-device running Large Language Models (LLMs) is nowadays a critical enabler towards preserving user privacy. We observe that the attention operator falls back from the special-purpose NPU to the general-purpose CPU/GPU because of quantization sensitivity in state-of-the-art frameworks. This fallback results in a degr...","url_abs":"https://arxiv.org/abs/2508.16703","url_pdf":"https://arxiv.org/pdf/2508.16703v4","authors":"[\"Wangsong Yin\",\"Daliang Xu\",\"Mengwei Xu\",\"Gang Huang\",\"Xuanzhe Liu\"]","published":"2025-08-22T07:41:35Z","proceeding":"cs.PF","tasks":"[\"cs.PF\",\"cs.AI\",\"cs.LG\"]","methods":"[\"Large Language Model\",\"Language Model\"]","has_code":false}