ClawArena: Benchmarking AI Agents in Evolving Information Environments

In three linesClawArena is a benchmark evaluating AI agents in evolving information environments. It tests agents' ability to maintain correct beliefs amid contradictory sources, dynamic updates, and implicit user preferences. 12 multi-turn scenarios, 337 evaluation rounds, 5 frameworks and 18 language models assessed.

Summary generated by Claude — human-verified