AI summaryⓘ
The authors compare two methods, SIGReg and VISReg, that shape how a model's internal representation works for planning tasks. They find that VISReg, which allows more control over shaping this space and uses larger batches, helps models plan better especially when facing new situations, while SIGReg does not. To understand why, they create a new score called VIScore that measures how well the model's components work together for successful planning. VIScore predicts planning success better than previous measures across different tasks and models, highlighting three key aspects: accuracy, influence, and simplicity in the model. The authors suggest VIScore can guide future improvements in designing world models for planning.
latent spaceisotropic Gaussianworld modelregularizationself-supervised learningout-of-domainpredictorplanning successSpearman correlationcalibration error
Abstract
Regulating the latent space to an isotropic Gaussian distribution provides a stable and information-maximized landscape for world model planning. However, the latent space property and successful planning remain disconnected. We first study this by comparing SIGReg and VISReg, two regularization loss functions with the same distribution target but different properties. Compared with SIGReg, VISReg has more flexibility in controlling the weights of center, scale, and shape regularization, and a larger batch size brings a finer distribution approximation. We find that the former, despite being beneficial in self-supervised learning (SSL), does not help the planning, whereas the latter improves the planning success on out-of-domain (OOD) datasets. This motivates a deep understanding of the factors that correlate with the success rate. Unlike the previous metrics focusing on the encoded latent only, we propose the Veracity-Influence-Sobriety score (VIScore), a metric that quantifies the reachability and capacity of a predictor given the encoded feature, and the hallucination of the searching-based planner. Compared with straightness, physical-state probing, and empowerment, we show that, with the measurement covering encoder, predictor, and planner, VIScore explains the success rate better than the others, as reflected by a strong Spearman correlation. Specifically, VIScore consistently achieves a Spearman correlation over 0.75 on both seen and unseen models and datasets on the cross-task success rate pool. Moreover, VIScore is the only metric that has a calibration error below the constant fit across all testing scenarios, showcasing the importance of these three aspects in planning success. We hope this metric can help future studies on world model design and diagnosis.