Research · MarkTechPost ·
Perplexity AI Releases WANDR: An Open Benchmark Evaluating Research Agents That Must Search Wide And Deep
Perplexity released WANDR, an open benchmark and evaluation harness for research agents, featuring 500 evidence-heavy tasks. It measures broad entity discovery and cited, re-verifiable evidence; Perplexity Search as Code scored 0.363 soft F1 and 0.133 hard F1.