CodeARC: Benchmarking Reasoning Capabilities of LLM Agents for Inductive Program Synthesis Paper • 2503.23145 • Published Mar 29, 2025 • 35