为无线大模型设计早退出机制，加速推理并提升未知任务表现
Fast Wireless Foundation Models with Early-Exits
- 在冻结的无线大模型编码器中添加轻量级任务头，按需选择中间层输出
- 推理耗时减少最多93% FLOPs，且对未见任务的准确率反而更高
- 固定任务对应单一出口比动态难度路由更有效，适合实际部署
尽管无线基础模型（FMs）展现出推动6G智能网络的潜力，其高计算成本仍是部署的主要障碍。现有模型对所有任务均执行全深度推理，既低效又可能降低对未知分布任务的性能。本文提出一种新型早退出基础模型框架，在冻结的无线编码器中为每个任务配置轻量级头部，并在最合适的层级输出，实现按任务定制的可变深度推理。实验表明，中间层特征不仅显著加速推理（最高减少93% FLOPs），还能提供更具迁移能力的表示，使未见任务的准确率超过完整编码器。此外，我们发现针对每个任务采用固定出口策略，比基于样本难易度动态分配出口更有效。
原文摘要 · Abstract (English)
While wireless foundation models (FMs) are demonstrating strong potential to enable AI-Native 6G networks, their high computational cost remains a critical barrier to deployment. The large computational cost stems from the rigid, full-depth execution of the FM backbone for every task, a process we show is not only inefficient but can also degrade performance on unseen out-of-distribution (OOD) tasks. In this paper, we propose a novel early-exit FM framework that attaches lightweight, per-task heads, at the most appropriate exit-stage of a frozen wireless FM encoder, enabling variable-depth inference tailored to each task's preferred representation depth. Our results demonstrate that these intermediate-layer features not only speed-up inference significantly (up to 93% fewer FLOPs), but also provide more transferable representations that exceed the full encoder accuracy on unseen tasks. We further demonstrate that a simple fixed-exit strategy per task is more effective than traditional early-exiting policies that route different samples to different exits based on their perceived difficulty levels.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。