Two papers accepted at NeurIPS 2023: on how to scale knowledge graph embeddings and treat them as generative models (as oral!) and on formalizing, understanding and mitigating reasoning shortcuts in neuro-symbolic AI.