English(EN) Compute-in-Memory Attention: A Time-Domain Analog Softmax Circuit with RC-Tunable Temperature

发布了用于内存计算加速器的新型模拟Softmax电路

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-07 04:00

研究人员开发了一种新颖的模拟Softmax电路，专为内存计算（CIM）加速器设计，旨在减少Transformer注意力机制的计算开销。该电路直接在CIM生成的电压分数上运行，将其转换为时域事件，从而在没有中间数模转换的情况下产生指数权重。该设计通过调整斜率和RC时间常数，可以实现可编程的有效温度。已使用MemTorch在硬件感知Transformer模型中进行了模拟和集成，验证损失在理想Softmax基线的2.5%以内。 AI

影响有望降低注意力机制AI硬件加速器的功耗和延迟。

排序理由详细介绍用于AI计算的新型硬件电路的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 32 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 详细介绍用于AI计算的新型硬件电路的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, infra; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Ankur Singh, Ashish Gautam, Shruti R. Kulkarni, Guojing Cong · 2026-09-07 04:00

内存计算注意力：一种具有RC可调温度的时域模拟Softmax电路

arXiv:2609.04266v1 Announce Type: cross Abstract: Softmax is a key operation in Transformer attention, but its exponentiation and normalization add significant overhead in compute-in-memory (CIM) accelerators, especially when analog attention scores must first be converted to the…

本文如何被排名

报道来源 [1]

内存计算注意力：一种具有RC可调温度的时域模拟Softmax电路

相关实体

相关话题