{"ID":5551775,"CreatedAt":"2026-07-02T01:54:51.863792489Z","UpdatedAt":"2026-07-04T09:53:58.593020999Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.00712","arxiv_id":"2607.00712","title":"Towards Memory-Efficient Autoregressive Video Generation via Instance-Specific Parametric Absorption","abstract":"Autoregressive (AR) streaming models have emerged as a powerful paradigm for long video generation. However, the linearly growing Key-Value (KV) cache poses a significant bottleneck, leading to memory overload and degraded inference throughput. A common compression method is to drop redundant KV tokens, which often breaks long-range dependencies, resulting in temporal flickering and identity loss. In this paper, we propose Instance-Specific Parametric Absorption (ISPA), a novel framework that shifts the KV cache compression from discarding to distilling. The core idea is to transit a subset of layers from Full-Attention (F-Layers) to memory-efficient Local-Attention (L-Layers) by \"absorbing\" historical context into the model's weights. Specifically, during a brief warmup phase, ISPA monitors the output discrepancy between global and local attention. At the transition point, we solve a closed-form least-squares problem to compute an instance-specific weight modulation that compensates for the missing history. Experiments across architectures (1.3B to 14B) demonstrate that ISPA can remove up to 50\\% of the KV cache with near-lossless visual quality. We hope this perspective encourages future work to explore parametric memory consolidation beyond external token-level cache management for streaming generative models.","short_abstract":"Autoregressive (AR) streaming models have emerged as a powerful paradigm for long video generation. However, the linearly growing Key-Value (KV) cache poses a significant bottleneck, leading to memory overload and degraded inference throughput. A common compression method is to drop redundant KV tokens, which often bre...","url_abs":"https://arxiv.org/abs/2607.00712","url_pdf":"https://arxiv.org/pdf/2607.00712v1","authors":"[\"Xiaomeng Fu\",\"Jia Li\",\"Yiming Hu\",\"Yong Wang\",\"Hayden Kwok-Hay So\",\"Jiao Dai\",\"Xiangxiang Chu\",\"Jizhong Han\"]","published":"2026-07-01T09:59:28Z","proceeding":"cs.CV","tasks":"[\"cs.CV\",\"cs.MM\"]","methods":"[]","has_code":false}