Skip to content

Drift-Bench

"Drift-Bench: Cooperative Breakdowns in Conversational AI." arXiv:2602.02455, 2026.

Key findings used in wiki

  • Introduces a benchmark specifically designed to measure cooperative breakdowns in multi-turn dialogue
  • Demonstrates that conversational AI systems fail to maintain cooperative alignment over extended interactions
  • Breakdowns occur even in non-adversarial settings where users are cooperative
  • Complements PBSuite by showing that failure is not limited to adversarial scenarios
  • Provides evaluation methodology that influenced InvisibleBench's cooperative breakdown scoring