sn-dimension-research

Cross-validation in deep research involves filtering evidence across academic, code, social, and general web sources, verifying original sources, and documenting the reasoning for keeping or discarding findings to prevent overconfident summaries.

Generating an evidence-backed sub-report for a single dimension requires executing multi-round searches based on a predefined search strategy, checking evidence sufficiency, and writing the traceable results into a dedicated markdown file.

Before running single-dimension research workflows, you need a report directory containing a request file and a plan file that defines the dimension, key questions, search strategy, expected output, and optional cross-dimension constraints.

Evidence sufficiency control checks if enough verified evidence exists, writes the sub-report markdown file, and updates the dimension status to done in the plan file only when evidence is present.

Yes, you can route search strategies across academic, code, social, and general web sources by defining the routing within the plan file, allowing the research process to respect dimension-specific source preferences.