Neuro-symbolic (NeSy) AI improves deep learning by integrating reasoning, prior knowledge, and constraints, making it in theory ideal for high-stakes applications. However, this promise depends on learning high-quality abstractions, which is challenging due to potential reasoning shortcuts. This project explores this problem in depth.
explainable-ai concept-learning neuro-symbolic-ai reasoning-shortcuts deepproblog mnist-addition distributed-alignment-search causal-abstraction
-
Updated
Apr 13, 2025 - Python