search-web

You can structure web research by decomposing goals into query clusters and running parallel searches. This evaluator-optimizer loop logs results and assesses quality to ensure findings are traceable and reproducible.

An evaluator-optimizer loop for web research is a mechanism that iteratively improves search findings. It applies criterion-based evaluation covering coverage, sources, consistency, specificity, and recency to guide query optimization.

To evaluate multi-round search findings, apply structured criteria such as coverage, source consistency, specificity, and recency. These evaluation metrics guide iterative query optimization until quality requirements are satisfied.

Structured logging for search traceability is needed when consolidating research deliverables. It ensures sources are tracked and results are reproducible, providing clear handoff criteria to decide when to complete the research.

Yes, you can use parallel subagent searches to decompose a research goal. This approach breaks down the overarching objective into actionable components and generates structured search queries to gather findings simultaneously.