{"ID":2826296,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2512.19445","arxiv_id":"2512.19445","title":"Sensitivity-Aware Mixed-Precision Quantization for ReRAM-based Computing-in-Memory","abstract":"Compute-In-Memory (CIM) systems, particularly those utilizing ReRAM and memristive technologies, offer a promising path toward energy-efficient neural network computation. However, conventional quantization and compression techniques often fail to fully optimize performance and efficiency in these architectures. In this work, we present a structured quantization method that combines sensitivity analysis with mixed-precision strategies to enhance weight storage and computational performance on ReRAM-based CIM systems. Our approach improves ReRAM Crossbar utilization, significantly reducing power consumption, latency, and computational load, while maintaining high accuracy. Experimental results show 86.33% accuracy at 70% compression, alongside a 40% reduction in power consumption, demonstrating the method's effectiveness for power-constrained applications.","short_abstract":"Compute-In-Memory (CIM) systems, particularly those utilizing ReRAM and memristive technologies, offer a promising path toward energy-efficient neural network computation. However, conventional quantization and compression techniques often fail to fully optimize performance and efficiency in these architectures. In thi...","url_abs":"https://arxiv.org/abs/2512.19445","url_pdf":"https://arxiv.org/pdf/2512.19445v1","authors":"[\"Guan-Cheng Chen\",\"Chieh-Lin Tsai\",\"Pei-Hsuan Tsai\",\"Yuan-Hao Chang\"]","published":"2025-12-22T14:44:05Z","proceeding":"cs.AR","tasks":"[\"cs.AR\",\"cs.ET\"]","methods":"[]","has_code":false}