What problem does it solve? It turns a fuzzy research idea into a concrete, implementation-ready method plan by freezing a Problem Anchor and iterating through structured external review until the proposal is focused, feasible, and top-venue ready. ## Core Features & Use Cases - Problem Anchor Freezing: Extracts the immutable bottom-line problem, constraints, and success conditions so every revision round stays on target and drift is explicitly flagged. - Iterative GPT-5.4 Review Loop: Sends proposals to a Codex-based reviewer across up to 5 rounds, scoring 7 dimensions (problem fidelity, method specificity, contribution quality, frontier leverage, feasibility, validation focus, venue readiness) until the score reaches 9. - Structured Logging: Saves every round's proposal, review, refinement, score history, and a final clean proposal under refine-logs/. - Use Case: A researcher has a vague idea like "improve long-video understanding with memory." The skill scans local papers, identifies the exact technical gap, drafts a minimal mechanism-level method, and iterates with GPT-5.4 until the plan is sharp enough to hand off to experiment planning. ## Quick Start Ask the assistant to refine your research plan by providing your problem and rough approach, for example: refine my approach on PROBLEM: long-video temporal grounding | APPROACH: add a memory module to the VLM.
