SOURCE-LINKED INTELLIGENCE

Sensory Precision Inference for Multimodal Arbitration under Uncertainty

arXiv · AI, language, vision and robotics · article · Sep 14, 2026 · UTC

Autonomous agents operating on multisensory data cannot assume that all sensory modalities remain consistently informative. In real environments, sensory streams are frequently corrupted by noise, missing data, or inter-modal incongruence, requiring adaptive arbitration between competing sensory hypotheses. While active inference provides a principled framework for uncertainty-guided inference, the role of dynamically inferred sensory precision in generative multimodal arbitration under sensory conflict remains comparatively underexplored. We propose a multimodal perceptual inference model in

Read original source ↗ Open in workspace

recordType

paper

region

Global