Evaluations

Each case runs the full research pipeline and is scored by an independent judge model on content, tool use, sources and reasoning.