Reduced Matrix Multiplication: Input-Adaptive Matrix-Product Reduction for LLM Inference

• 类型：arxiv
• 标识：2608.13426
• 链接：http://arxiv.org/abs/2608.13426v1
• 主分类：llm-infra
• 形态：method
• 被引：0
• 被引来源：Semantic Scholar
• S2被引：0
• 影响力被引：0
• TLDR：Reduced Matrix Multiplication is proposed, a training-free, input-adaptive inference method that reduces Transformer matrix products by selecting informative slices along their contraction dimensions, without modifying model weights, and it is shown that the same principle extends to multimodal vision-language inference.
• 待LLM分类：否
• 标题中文：缩减矩阵乘法：用于 LLM 推理的输入自适应矩阵乘积缩减
• TLDR中文：本文提出 Reduced Matrix Multiplication，一种免训练的输入自适应推理方法，通过沿收缩维度选取信息性切片来减少 Transformer 矩阵乘积，且不修改模型权重，并表明同一原理可扩展到多模态视觉-语言推理。
• 副分类：multimodal
• 来源文件：
• /inbox/tom/_candidates/2026-08-14-agent-rag-longcontext-candidates.json
• [S2 enrich]