model-review

# 数模方案评审 via Codex MCP (xhigh reasoning) 通过外部 LLM 以最高推理深度对建模方案进行多轮评审，模拟数模竞赛评委视角。 ## Constants - REVIEWER_MODEL = `gpt-5.4` — Model used via Codex MCP. Must be an OpenAI model (e.g., `gpt-5.4`, `o3`, `gpt-4o`) ## Context: $ARGUMENTS ## Prerequisites - **Codex MCP Server** configured in Claude Code: ```bash claude mcp add codex -s user -- codex mcp-server ``` - This gives Claude Code access to `mcp__codex__codex` and `mcp__codex__codex-reply` tools ## Workflow ### Step 1: 收集建模上下文 Before calling the external reviewer, compile a comprehensive briefing: 1. Read project documents (e.g., PROBLEM_BRIEF.md, MODEL_REPORT.md, PROBLEM_ANALYSIS.md, paper drafts) 2. Read any result files for key findings and solving history 3. Identify: core modeling approach, mathematical methods, key results, known weaknesses ### Step 2: Initial Review (Round 1) Send a detailed prompt with xhigh reasoning: ``` mcp__codex__codex: config: {"model_reasoning_effort": "xhigh"} prompt: | [完整建模方案 + 赛题描述 + 具体问题] 请以数模竞赛资深评委/阅卷专家的身份评审。按以下维度打分和评价: 1. 数学建模严谨性（模型假设是否合理、推导是否正确） 2. 方法创新性（是否有亮点、是否超越简单套用） 3. 结果可靠性（计算结果是否合理、是否有验证） 4. 论文规范性（结构是否完整、表述是否清晰） 请以国赛一等奖/美赛O奖标准严格评审。 ``` ### Step 3: Iterative Dialogue (Rounds 2-N) Use `mcp__codex__codex-reply` with the returned `threadId` to continue the conversation: For each round: 1. **Respond** to criticisms with evidence/counterarguments 2. **Ask targeted follow-ups** on the most actionable points 3. **Request specific deliverables**: model improvements, analysis suggestions, problem-method-result mapping Key follow-up patterns: - "If we change model assumption X to Y, does that change your assessment?" - "What's the minimum additional analysis to satisfy concern Z?" - "Please suggest the highest-impact improvements within the remaining competition time" - "Please write a mock competition review with scores for each dimension" - "Give me a problem-method-result mapping for each sub-problem" ### Step 4: Convergence Stop iterating when: - Both sides agree on the modeling approach and its validation requirements - A concrete improvement plan is established - The paper structure and narrative are settled ### Step 5: Document Everything Save the full interaction and conclusions to a review document in the project root: - Round-by-round summary of criticisms and responses - Final consensus on modeling approach, validation, and improvements - Problem-method-result mapping (问题-方法-结果对应表) - Prioritized TODO list with estimated time costs - Paper outline if discussed Update project memory/notes with key review conclusions. ## Key Rules - ALWAYS use `config: {"model_reasoning_effort": "xhigh"}` for reviews - Send comprehensive context in Round 1 — the external model cannot read your files - Be honest about weaknesses — hiding them leads to worse feedback - Push back on criticisms you disagree with, but accept valid ones - Focus on ACTIONABLE feedback — "what analysis or improvement would fix this?" - Document the threadId for potential future resumption - The review document should be self-contained (readable without the conversation) ## Prompt Templates ### For initial review: "我将展示一个完整的数学建模方案，请以数模竞赛资深评委的身份进行严格评审。按国赛一等奖/美赛O奖标准..." ### For improvement design: "请设计在剩余竞赛时间内能最大幅度提分的改进方案。当前时间余量: [describe]。请给出具体的改进步骤。" ### For paper structure: "请将建模方案转化为数模论文大纲，包含各章节的核心论点和图表规划。" ### For problem-method-result mapping: "请给出问题-方法-结果对应表：每个子问题使用什么方法、预期什么结果、实际得到什么结果。" ### For mock review: "请模拟数模竞赛评审，给出: 总体评价、数学严谨性评分、创新性评分、结果可靠性评分、论文规范性评分、改进建议、获奖预估等级。"