Opt-in Self-Check

What is Opt-in Self-Check?

We built a framework that analyzes LLM agent behavior during execution instead of only final outcomes. It detects failure signals such as: - tool errors - repetitive actions - stagnation - degraded exploration This helps understand how and when LLM agents begin to fail during runtime, not just after completion.