BenchmarkApril 11, 20261 min read
AI Models Fail Miserably at Asking for Help, Scoring a Dismal 17.5% on ProactiveBench
A new benchmark has revealed that AI models are woefully inept at requesting assistance when faced with incomplete or unclear information, with even the largest models scoring poorly. This shortcoming has significant implications for developers and users relying on AI for critical tasks.
ProactiveBench tests whether multimodal language models ask users for help when visual information is missing. Out of 22 models tested, almost none ask for what they need, but a simple reinforcement learning approach hints at a fix. The article AI models would rather guess than ask for help, researchers find appeared first on The Decoder.