MonitorBench: A Comprehensive Benchmark for Chain-of-Thought Monitorability in Large Language Models Paper • 2603.28590 • Published Mar 30 • 22
