NexusFlow：借助可逆流网络在部分监督下统一异构任务
NexusFlow: Unifying Disparate Tasks under Partial Supervision via Invertible Flow Networks
用可逆耦合层构造代理网络，把结构差异极大的任务（稠密地图重建与稀疏多目标跟踪）的潜变量特征映射到共享规范空间做对齐，避免表征坍缩；这套 plug-and-play 的对齐模块可直接用于视觉中标注不完整、任务异构的多任务训练，例如不同区域只标注不同任务的数据。
原文依据 NexusFlow introduces a set of surrogate networks with invertible coupling layers to align the latent feature distributions of tasks
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论部分监督多任务学习（PS-MTL）旨在标注不完整的情况下利用跨任务知识。然而，现有方法大多集中于同质稠密预测任务这一较简单的设定，尚未探索从结构上多样化的任务中学习这一更贴近现实的挑战。为此，我们提出 NexusFlow，一个新颖、轻量且即插即用的框架，在两种设定下均有效。NexusFlow 引入一组带有可逆耦合层的代理网络，用于对齐各任务的潜变量特征分布，从而构建统一表征以实现有效的知识迁移。耦合层是双射的，在将特征映射到共享规范空间的同时保留信息。这种可逆性避免了表征坍缩，并使得在结构不同的任务之间进行对齐成为可能，且不降低表达能力。我们首先在域划分的自动驾驶这一核心挑战上评估 NexusFlow：稠密地图重建与稀疏多目标跟踪在不同地理区域分别受到监督，从而同时造成结构差异与强烈的域间差距。NexusFlow 在 nuScenes 上取得新的最优结果，优于强部分监督基线。为证明通用性，我们进一步在 NYUv2 上使用分割、深度和表面法线这三个同质稠密预测任务测试 NexusFlow，作为具有代表性的 N 任务 PS-MTL 场景。NexusFlow 在所有任务上均带来一致增益，证实了其广泛适用性。
英文原文摘要 1572 字符
Partially Supervised Multi-Task Learning (PS-MTL) aims to leverage knowledge across tasks when annotations are incomplete. Existing approaches, however, have largely focused on the simpler setting of homogeneous, dense prediction tasks, leaving the more realistic challenge of learning from structurally diverse tasks unexplored. To this end, we introduce NexusFlow, a novel, lightweight, and plug-and-play framework effective in both settings. NexusFlow introduces a set of surrogate networks with invertible coupling layers to align the latent feature distributions of tasks, creating a unified representation that enables effective knowledge transfer. The coupling layers are bijective, preserving information while mapping features into a shared canonical space. This invertibility avoids representational collapse and enables alignment across structurally different tasks without reducing expressive capacity.We first evaluate NexusFlow on the core challenge of domain-partitioned autonomous driving, where dense map reconstruction and sparse multi-object tracking are supervised in different geographic regions, creating both structural disparity and a strong domain gap. NexusFlow sets a new state-of-the-art result on nuScenes, outperforming strong partially supervised baselines. To demonstrate generality, we further test NexusFlow on NYUv2 using three homogeneous dense prediction tasks, segmentation, depth, and surface normals, as a representative N-task PS-MTL scenario. NexusFlow yields consistent gains across all tasks, confirming its broad applicability.