发表机构
Chongqing Finance and Economics College（重庆财经学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究揭示编码器越强越依赖MMD对齐且越受重平衡损害，据此简化配方为冻结嵌入加单MMD，保持性能并提升未见域泛化。
AI 中文摘要
分布对齐何时能帮助冻结的基础模型嵌入在声学域间泛化？针对跨域蚊子物种分类，我们报告了一个强度单调定律：编码器在目标任务上越强，其未见域泛化越依赖于分布对齐（MMD）项，且越受域重平衡采样的损害。在四个编码器家族和编码器内HuBERT层扫描（n=8）中，重平衡分支与编码器强度完全有序（Spearman -1.000），而MMD收益分支在每个流内单调，合并后为-0.857；固定架构仅改变表示强度会将重平衡效应从收益翻转为崩溃。该定律具有可操作性：单个MMD项是强编码器上的唯一杠杆，因此我们将该领域的默认配方简化为冻结的Perch 2.0嵌入、轻量探针、交叉熵、一个MMD和输入增强。简化配方保持在完整组合的种子噪声范围内（BA_unseen 0.299±0.006 vs. 0.307±0.014）。作为同一定律的边界条件，三个社区默认设置（骨干微调、多模态融合和域重平衡）在留域协议下均损害未见域准确率，通过单变量、多种子证据展示。我们提出一种机制及其解释的配方，而非排行榜条目。
英文摘要
When does distribution alignment help a frozen foundation-model embedding generalize across acoustic domains? For cross-domain mosquito-species classification we report a strength-monotonic law: the stronger an encoder is on the target task, the more its unseen-domain generalization relies on a distribution-alignment (MMD) term, and the more it is harmed by domain-rebalanced sampling. Across four encoder families and a within-encoder HuBERT layer sweep (n=8), the rebalancing leg orders exactly with encoder strength (Spearman -1.000), while the MMD-benefit leg is monotonic within each stream and -0.857 pooled; fixing architecture and varying only representation strength flips the rebalancing effect from benefit to collapse. The law is actionable: a single MMD term is the sole lever on a strong encoder, so we reduce the field's default recipe to a frozen Perch 2.0 embedding, a lightweight probe, cross-entropy, one MMD, and input augmentation. The reduced recipe stays within seed noise of the full composite (BA_unseen 0.299+/-0.006 vs. 0.307+/-0.014). As boundary conditions of the same law, three community defaults (backbone fine-tuning, multi-modal fusion, and domain rebalancing) each hurt unseen-domain accuracy under a leave-domain protocol, shown with single-variable, multi-seed evidence. We present a mechanism and the recipe it explains, not a leaderboard entry.
