Abstraction and Reasoning Corpus (ARC)
Emerging9papers using it
2021first seen
The Abstraction and Reasoning Corpus (ARC) is a benchmark that contains tasks designed to evaluate the abstract reasoning and generalization capabilities of models, particularly in the context of program synthesis.
Papers using Abstraction and Reasoning Corpus (ARC) (9)
- MADIL: An MDL-based Framework for Efficient Program Synthesis in the ARC
BenchmarkFrom Reasoning to Generalization: Knowledge-Augmented LLMs for ARC BenchmarkANPL: Towards Natural Programming with Interactive DecompositionFast and flexible: Human program induction in abstract reasoning tasksCodeIt: Self-Improving Language Models with Prioritized Hindsight ReplayNeural-guided, Bidirectional Program Search for Abstraction and
ReasoningProgram Synthesis using Inductive Logic Programming for the Abstraction and Reasoning CorpusConceptSearch: Towards Efficient Program Search Using LLMs for
Abstraction and Reasoning Corpus (ARC)Communicating Natural Programs to Humans and Machines