Urgent.News

What's breaking now, across thousands of outlets.

More in AI

Stop Trusting Text-Only Agent Leaderboards: Lessons from Cua-Bench and Factorio

Stop Trusting Text-Only Agent Leaderboards: Lessons from Cua-Bench and Factorio Overfitting to leaderboard benchmarks has created a field of agents tuned for text puzzles but fragile the moment real…

  • Text-only agent leaderboards can be misleading
  • Cua-Bench and Factorio LE expose this fragility
  • Agents fail at stateful context, error recovery

More from Wednesday 26 August →