Act As a Real Researcher: A Suite of Benchmarks Evaluating Frontier LLMs and Agentic Harnesses in Research Lifecycle

In three linesAARRI-Bench evaluates AI agents on granular scientific research tasks. Even the best configuration (Mini-SWE-Agent + Claude Opus 4.7) achieves 68.3% success rate, revealing gaps in nuanced judgment and research ethics. The benchmark targets ability to emulate human researcher professionalism beyond macro-level execution.

Summary generated by Claude — human-verified