研究人员开发了几种新的方法来压缩和优化大型语言模型（LLMs），以提高效率并降低计算成本。SparseForge 通过优化稀疏掩码来专注于高效的半结构化稀疏化，以显著更少的重新训练 token 实现高精度。FASQ 引入了灵活的加速子空间量化，能够在没有校准数据的情况下实现连续的压缩级别，并在商品 GPU 上在准确性和速度方面均优于现有方法。此外，CoSpaDi 使用校准引导的稀疏字典学习进行结构化分解，改善了精度-压缩权衡。另一种方法 SplitZip 为分离式 LLM 服务提供了超快速的无损 KV 缓存压缩，显著加快了模型组件之间的数据传输速度。 AI
影响 这些在 LLM 压缩和优化方面的进展可能导致在功能较弱的硬件上更有效地部署大型模型，并缩短推理时间。
排序理由 arXiv 上发表了多篇研究论文，详细介绍了 LLM 压缩和优化的新颖方法。
AI 生成摘要 · Google Gemini · 来自 6 个来源。 我们如何撰写摘要 →
arXiv:2605.06402v1 Announce Type: new Abstract: Semi-structured sparsity provides a practical path to accelerate large language models (LLMs) with native hardware support, but post-training semi-structured pruning often suffers from substantial quality degradation due to strong s…
Semi-structured sparsity provides a practical path to accelerate large language models (LLMs) with native hardware support, but post-training semi-structured pruning often suffers from substantial quality degradation due to strong structural coupling. Existing methods rely on lar…
arXiv:2605.04084v1 Announce Type: new Abstract: Compressing large language models (LLMs) for deployment on commodity GPUs remains challenging: conventional scalar quantization is limited to fixed bit-widths (e.g., 8/4/3-bit), offers only a few discrete compression points, and typ…
arXiv:2509.22075v5 Announce Type: replace Abstract: Post-training compression of large language models (LLMs) often relies on low-rank weight approximations that represent each column of the weight matrix in a shared low-dimensional subspace. This strategy is computationally effi…
arXiv:2605.01255v1 Announce Type: new Abstract: Training large language models (LLMs) is highly memory-intensive, as training must store not only weights and optimizer states but also intermediate activations for backpropagation. While existing memory-efficient methods largely fo…
arXiv:2605.01708v1 Announce Type: cross Abstract: Contemporary systems serving large language models (LLMs) have adopted prefill-decode disaggregation to better load-balance between the compute-bound prefill phase and the memory-bound decode phase. Under this design, prefill work…