AI 中文总结
针对时间序列基础模型在实例归一化骨干上路由崩溃的问题，提出原始路由混合适配器（RR-MoA），通过因果干预在归一化前输入上路由，在冻结骨干下显著优于多种适配方法，并具有跨骨干泛化性。
AI 中文摘要
时间序列基础模型（TSFMs）通常通过将单个可训练头部附加到冻结骨干上来适应新数据，这是一种一刀切的设置，无法充分拟合异构机制。将头部替换为专家混合是标准的升级方式，但在实例归一化骨干（主导的TSFM设计类别）上，该方式失败：路由熵降至零，一个专家吸收所有输入，我们将这种失败称为归一化诱导的路由崩溃。标准的MoE救援机制无法修复此问题，因为原因在于路由器的输入，而非其优化过程。编码器前的归一化剥离了路由器区分不同机制所需的统计信息。互信息分解使这一点更加精确，并产生一个信号比率，该比率在训练前计算，可预测数据集脆弱性（Spearman ρ = -0.88）。包括视觉模态复制在内的八项因果对照实验，将实例归一化确定为原因。解决方案是一种最小因果干预：原始路由混合适配器（RR-MoA），它在原始、归一化前的输入上进行路由。在严格冻结骨干下，RR-MoA在54/54次比较中胜过最强的固定适配器，并显著优于LoRA、TRACE、AdaMix和全参数微调。该效果在六个骨干和一个插补任务上具有泛化性。冻结的RR-MoA还比全参数微调高出12-79%（冻结悖论）；两个架构不同的变体证实了该原理超越了这一特定路由器。
英文摘要
Time series foundation models (TSFMs) commonly adapt to new data by attaching a single trainable head to a frozen backbone, a one-size-fits-all setup that underfits heterogeneous regimes. Replacing the head with a mixture of experts is the standard upgrade, but on instance-normalized backbones (the dominant TSFM design class) it fails: routing entropy collapses to zero and one expert absorbs every input, a failure we call normalization-induced routing collapse. Standard MoE rescue mechanisms do not repair it, because the cause is in the router's input, not its optimization. Pre-encoder normalization strips the statistics a router would need to tell regimes apart. A mutual-information decomposition makes this precise and yields a signal-ratio that, computed before training, predicts dataset vulnerability (Spearman $ρ= -0.88$). Eight causal controls, including a vision-modality replication, isolate instance normalization as the cause. The prescription is a minimal causal intervention: Raw-Routed Mixture of Adapters (RR-MoA), which routes on the raw, pre-normalization input. Under a strictly frozen backbone, RR-MoA wins 54/54 comparisons against the strongest fixed adapter and significantly outperforms LoRA, TRACE, AdaMix, and full fine-tuning. The effect generalizes across six backbones and an imputation task. Frozen RR-MoA also beats full fine-tuning by 12-79% (the Frozen Paradox); two architecturally distinct variants confirm the principle generalizes beyond this specific router.
CommentsAccepted at NeurIPS 2026 (poster). 50 pages, 9 figures
