research-refine

You can refine a vague research direction by iteratively reviewing a problem statement and approach. This process freezes an immutable Problem Anchor to ensure focus, then uses GPT-5.5 review to evaluate problem fidelity, method specificity, contribution quality, and frontier technique leveraging.

Iterative review in research methodology repeatedly evaluates a problem statement against fidelity, specificity, contribution, and frontier leverage criteria. It uses GPT-5.5 to review the research approach across multiple rounds until a specified threshold score is met.

To use foundation-model techniques in research design, the refinement process evaluates whether your approach appropriately leverages LLMs, VLMs, Diffusion, RL, distillation, or inference-time scaling to address the core bottleneck with the smallest effective intervention.

Yes, you can set both a maximum number of review rounds and a quality threshold score. The refinement process stops when either the maximum round limit is reached or the review score meets your specified threshold.

A Problem Anchor is an immutable frozen statement of your core research issue. It is needed to maintain focus during iterative refinement, ensuring that all methodology adjustments and mechanism additions directly serve the original problem without scope drift.

Your research approach may lack specificity and contribution quality if it is not anchored to a frozen problem or evaluated against modern frontier techniques. Iterative GPT-5.5 review identifies these gaps by assessing the minimal mechanism needed to fix the core bottleneck.