Perplexity cuts tool-call failures 21.2% by training on user mistakes
Perplexity’s Computer model reduced tool-call failures by 21.2% against an earlier trained checkpoint, giving real-world agent training a measurable payoff. The method keeps useful steps from successful sessions, then turns user corrections and tool errors into short, grounded hints that teach the model how to fix bad actions. It runs after reinforcement learning in synthetic environments, using opted-in, PII-free sessions to expose failures those environments miss.