NSA: Neuro-symbolic ARC Challenge
This addresses the problem of general reasoning in AI, which is difficult for both machine learning and search methods, by providing a novel approach that significantly boosts performance on a key benchmark.
The paper tackles the Abstraction and Reasoning Corpus (ARC) challenge, which tests general reasoning abilities, by proposing a neuro-symbolic method that combines a transformer for proposal generation with combinatorial search, achieving a 27% improvement over comparable state-of-the-art on the ARC evaluation set.
The Abstraction and Reasoning Corpus (ARC) evaluates general reasoning capabilities that are difficult for both machine learning models and combinatorial search methods. We propose a neuro-symbolic approach that combines a transformer for proposal generation with combinatorial search using a domain-specific language. The transformer narrows the search space by proposing promising search directions, which allows the combinatorial search to find the actual solution in short time. We pre-train the trainsformer with synthetically generated data. During test-time we generate additional task-specific training tasks and fine-tune our model. Our results surpass comparable state of the art on the ARC evaluation set by 27% and compare favourably on the ARC train set. We make our code and dataset publicly available at https://github.com/Batorskq/NSA.