Out-of-Distribution Generalization in the ARC-AGI Domain: Comparing Execution-Guided Neural Program Synthesis and Test-Time Fine-Tuning
This addresses the problem of compositional generalization in AI for open-world domains like ARC-AGI, but the findings appear incremental as they compare existing methods without introducing a new paradigm.
The paper tackled out-of-distribution generalization in the ARC-AGI domain by comparing execution-guided neural program synthesis and test-time fine-tuning, finding that the synthesis approach outperformed others in composing novel solutions, with test-time fine-tuning mainly eliciting in-distribution knowledge.
We run a controlled compositional generalization experiment in the ARC-AGI domain: an open-world problem domain in which the ability to generalize out-of-distribution is, by design, an essential characteristic for success. We compare neural program synthesis and test-time fine-tuning approaches on this experiment. We find that execution-guided neural program synthesis outperforms all reference algorithms in its ability to compose novel solutions. Our empirical findings also suggest that the success of TTFT on ARC-AGI lies mainly in eliciting in-distribution knowledge that the LLM otherwise fails to rely on directly.