Conjecture—Compositional representations required for specifying models and tasks
Establish the necessity of compositional representation frameworks for specifying embodied agents’ world models and tasks, enabling generalization and rapid adaptation to novel environments.
References
We conjecture that a specific form of compositionality is required of the representation that is absent from many architectures, and we address this conjecture in section \ref{sec:logic}.
— From Machine Learning to Robotics: Challenges and Opportunities for Embodied Intelligence
(2110.15245 - Roy et al., 2021) in Section 1 (Introduction: Guiding Questions)
However, the relationship between disentanglement and compositional generalization remains unclear.
— TPR-Attention for Combinatorial Generalization
(2608.30124 - Civelekoğlu et al., 31 Aug 2026) in Section 2, Related Work
The question is whether grounding, relational reasoning, and action knowledge can compose into reusable spatial expertise, supporting new tasks with less task-specific supervision.
— Anvil: Scaling Embodied Spatial Intelligence through Knowledge Recirculation
(2610.21882 - Ruan et al., 15 Oct 2026) in Section 'Limitations and Future Work', paragraph 'From specialist consolidation to compositional spatial competence'