MADIL uses MDL-based decomposition and composition of grids to synthesize interpretable programs from a few examples, solving 57 of 400 ARC evaluation tasks with an 82% generalization rate.
Chollet, A definition of intelligence for the real world, Journal of Artificial General Intelligence 11 (2) (2020) 27–30
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
MADIL: An MDL-based Framework for Efficient Program Synthesis in the ARC Benchmark
MADIL uses MDL-based decomposition and composition of grids to synthesize interpretable programs from a few examples, solving 57 of 400 ARC evaluation tasks with an 82% generalization rate.