CodeARC: Benchmarking Reasoning Capabilities of LLM Agents for Inductive Program Synthesis Paper • 2503.23145 • Published Mar 29 • 34