3701. On Evaluating Understanding and Generalization in the ARC Domain
Melanie Mitchell discusses the evaluation of understanding and generalization in machines using the Abstraction and Reasoning Corpus, highlighting challenges and a new benchmark called ConceptARC.