用对比反向动力学防止模型坍缩，无需预设高斯分布。
No Gaussian Required: Contrastive Inverse Dynamics for JEPA World Models
- 通过动作对比任务训练反向动态头，让隐变量区分动作
- 在多物体场景中比传统方法提升20-24个百分点，达80%
- 无需目标网络、重建损失或预训练，适合强化学习与规划任务
联合嵌入预测架构（JEPAs）通过预测未来嵌入来学习世界模型，但其目标存在恒定编码器的平凡解，因此实际系统需加入防坍缩机制（LeCun, 2022；Assran等, 2023；Bardes等, 2022；2024）。LeWorldModel（LeWM）采用SIGReg正则化，强制隐变量分布匹配各向同性高斯，以规定表征形态，独立于环境。本文提出，防坍缩信号可来自过渡数据本身。动作对比掩码过渡建模（AC-MTM）保留LeWM的前向隐变量预测目标，并引入仅用于训练的反向动力学头，使用动作NCE进行训练：每个隐状态转移必须从批次中其他动作里识别出产生它的动作，这是一个坍缩编码器无法完成的判别任务。训练后丢弃反向分支，测试时编码、前向预测、规划与计算仍与LeWM一致。在四个标准像素控制任务上，AC-MTM从零开始稳定训练，平均性能与SIGReg相当。在更难的多对象OGBench视觉场景任务中，结果表明预设几何成为瓶颈：AC-MTM达到80.0±2.0%成功率，优于SIGReg的58.0±2.0%，每种子实验提升20-24点。单次50步随机策略运行给出52%基线估计。对比反向动力学提供了无需分布假设的防坍缩信号，不依赖目标网络、停止梯度、预训练编码器或重构目标，我们还分析了该方法成立的动作空间与可观测性假设。代码已开源：https://github.com/jackboyla/action-contrastive-jepa
原文摘要 · Abstract (English)
Joint-Embedding Predictive Architectures (JEPAs) learn world models by predicting future embeddings, but the objective admits a trivial solution of a constant encoder, so every practical system adds an anti-collapse mechanism (LeCun, 2022; Assran et al., 2023; Bardes et al., 2022; 2024). LeWorldModel (LeWM) prevents collapse with SIGReg, a regularizer that forces the latent distribution to match an isotropic Gaussian: the representation is stabilized by prescribing what it must look like, independently of the environment it models. We argue that the anti-collapse pressure can instead come from the transition data itself. Action-Contrastive Masked Transition Modeling (AC-MTM) keeps LeWM's forward latent-prediction objective and adds a training-only inverse-dynamics head trained with Action-NCE: each latent transition must identify the action that produced it among the other actions in the batch, a discrimination task that a collapsed encoder provably fails. The inverse branch is discarded after training, leaving test-time encoding, forward prediction, planning, and compute identical to LeWM. On four standard pixel-control tasks under a matched planning protocol, AC-MTM trains stably from scratch and matches SIGReg on average. On the harder multi-object OGBench Visual Scene task, results are consistent with the prescribed geometry becoming a bottleneck: AC-MTM reaches 80.0$\pm$2.0% success versus 58.0$\pm$2.0% for SIGReg, improving by 20-24 points in each training seed. A single 50-episode random-policy run gives a 52% baseline estimate. Contrastive inverse dynamics thus provides a distribution-free anti-collapse signal that requires no target network, stop-gradient, pretrained encoder, or reconstruction objective, and we characterize the action-space and observability assumptions under which it holds. We make our code available at https://github.com/jackboyla/action-contrastive-jepa
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。