/refine

Iteratively improving a research paper draft involves a multi-round review-driven revision loop that applies fixes in place until a target quality score is reached. It extracts actionable items from each review to systematically close content gaps.

Detecting and fixing content gaps in an experiment plan is best handled by a multi-round review process that extracts actionable items and applies targeted revisions. This loop continues until the artifact reaches the target review score or convergence.

Automated content revision works on experiment plans and methods, not just paper drafts. This artifact-agnostic refinement loop applies to any research artifact referenced by wiki slugs or file paths until the target score is achieved.

Iterative research review leverages a wiki graph to update relevant entities and rebuild derived context edges after applying knowledge changes. This ensures wiki-backed context remains accurate after each revision loop.

The iterative improvement loop terminates when the target review score is reached, the review verdict is ready, or convergence and max rounds occur. It then outputs a final report containing the score trajectory and any unresolved issues.