深入剖析免训练低秩压缩中校准与截断误差的传播机制，为LLM高效压缩提供理论支撑。
arXiv:2608.08506v1 Announce Type: new Abstract: Training-free low-rank compression frameworks have been gaining prominence for LLM compression given t…
无需训练即可实现可变率KV缓存压缩，显著降低长上下文LLM推理内存开销，附大量实验验证效果。
arXiv:2607.15498v1 Announce Type: cross Abstract: The key-value (KV) cache is the main memory bottleneck in long-context large language model (LLM) in…
华为昇腾NPU上实现免训练4-bit FlashAttention，大幅降低LLM推理内存与计算开销。
arXiv:2607.04302v1 Announce Type: cross Abstract: We present HiFA4, a post-training operator-level design that executes both QK^T and PV in FlashAtten…
无需训练，推理时直接干预logit实现语言模型可控生成，SWAI方法简洁高效。
arXiv:2601.10960v2 Announce Type: replace-cross Abstract: Controllable generation requires language models to realize output characteristics such as r…
LLM长上下文推理新突破：无需额外训练即可精准估计KV缓存重要性，大幅降低推理开销。
arXiv:2605.27740v1 Announce Type: new Abstract: Long-context inference in large language models (LLMs) is bottlenecked by the linear growth of the sel…
无需额外训练，通过空间-时间池化与网格化巧妙提升视频大语言模型视觉token表征，ICLR 2026接收！
arXiv:2605.22078v1 Announce Type: cross Abstract: Recent advances in Multimodal Large Language Models (MLLMs) have significantly advanced video unders…
用字形先验和注意力引导语义混合，无需训练即可高质量渲染图像中被遮挡的文字，极具创新性。
arXiv:2605.16810v1 Announce Type: new Abstract: We present a training-free framework for occluded text rendering with a pretrained FLUX.1-dev backbone…
VLA模型存在动力学盲点？这篇论文提出无需训练的步速与路径校正方法，突破单帧观测局限。
arXiv:2605.11459v2 Announce Type: replace-cross Abstract: Vision-Language-Action (VLA) models achieve remarkable flexibility and generalization beyond…