Are LLMs Ready for Scientific Discovery? A Capability-Oriented Benchmark for AI Scientists Paper • 2607.11079 • Published 3 days ago • 4