Concept modulation models connect identifiability and extrapolation in conditional generative models. Attributes determine modulators, which govern latent concepts and ultimately observed features. The framework uses attribute potentials to characterize both the latent transformations compatible with observed data and the conditions under which agreement extends to unseen attributes. This provides algebraic extrapolation criteria and a common basis for several existing results in nonlinear ICA, causal representation learning, and perturbation modeling.
Dec 16, 2026
The Conditional Informer predicts ship trajectories by conditioning vessel dynamics on environmental information. Its attention mechanism lets vessel states query contextual variables such as weather, reflecting the directional dependence of motion on environmental conditions. A modality masking strategy improves robustness when contextual sensors are unavailable. Experiments combining AIS trajectories with ERA5 weather data show improved forecasting accuracy and substantially reduced errors when context is missing.
Jul 29, 2026
Concept extraction methods, including sparse autoencoders and transcoders, seek symbolic concepts within learned representations. This work studies their theoretical guarantees by treating concept extraction as a generative-model identification problem. A general identifiability theorem reduces the analysis to characterizing the intersection of two sets. Applying the framework to several common methods simplifies existing arguments and provides a foundation for designing concept extraction approaches with provable guarantees.
Apr 27, 2026
Jun 7, 2025
May 7, 2025
May 7, 2025
Apr 4, 2025
Mar 7, 2024
Apr 11, 2022
Apr 11, 2022