https://cygnusyang.github.io/posts/01-ai-%E5%B7%A5%E5%85%B7%E4%B8%8E%E6%99%BA%E8%83%BD%E4%BD%93/08-academic-research-skills/06-reviewer/技术 × 产品 × 机器人Hugo 0.160.1 & FixIt v0.4.6-20260517062719-ed8ec5c2zh-CNruohuyang@163.com (Cygnus Yang)ruohuyang@163.com (Cygnus Yang)Sat, 23 May 2026 00:00:00 +0800https://cygnusyang.github.io/posts/01-ai-%E5%B7%A5%E5%85%B7%E4%B8%8E%E6%99%BA%E8%83%BD%E4%BD%93/08-academic-research-skills/06-reviewer/12-7-reviewer-agents/Sat, 23 May 2026 00:00:00 +0800ruohuyang@163.com (Cygnus Yang)https://cygnusyang.github.io/posts/01-ai-%E5%B7%A5%E5%85%B7%E4%B8%8E%E6%99%BA%E8%83%BD%E4%BD%93/08-academic-research-skills/06-reviewer/12-7-reviewer-agents/<p>Academic Paper Reviewer 是 ARS 的&quot;质检部&quot;——它模拟真实学术期刊的多视角同行评议过程，用 7 个角色分工明确的 Agent 对论文进行结构化审稿。</p> <p>与传统&quot;一个 AI 读论文然后给意见&quot;不同，ARS 的审稿是<strong>多视角对抗式的</strong>：不同 Reviewer 有不同的专业侧重和批判立场，他们在论文上的分歧必须在综合报告中得到保留和讨论，而非被平均成一个&quot;综合评分&quot;。</p> <h2 class="heading-element" id="agent-全景"><span>Agent 全景</span> <a href="#agent-%e5%85%a8%e6%99%af" class="heading-mark"> <svg class="octicon octicon-link" viewBox="0 0 16 16" version="1.1" width="16" height="16" aria-hidden="true"><path d="m7.775 3.275 1.25-1.25a3.5 3.5 0 1 1 4.95 4.95l-2.5 2.5a3.5 3.5 0 0 1-4.95 0 .751.751 0 0 1 .018-1.042.751.751 0 0 1 1.042-.018 1.998 1.998 0 0 0 2.83 0l2.5-2.5a2.002 2.002 0 0 0-2.83-2.83l-1.25 1.25a.751.751 0 0 1-1.042-.018.751.751 0 0 1-.018-1.042Zm-4.69 9.64a1.998 1.998 0 0 0 2.83 0l1.25-1.25a.751.751 0 0 1 1.042.018.751.751 0 0 1 .018 1.042l-1.25 1.25a3.5 3.5 0 1 1-4.95-4.95l2.5-2.5a3.5 3.5 0 0 1 4.95 0 .751.751 0 0 1-.018 1.042.751.751 0 0 1-1.042.018 1.998 1.998 0 0 0-2.83 0l-2.5 2.5a1.998 1.998 0 0 0 0 2.83Z"></path></svg> </a> </h2><table> <thead> <tr> <th>Agent</th> <th>角色</th> <th>职责</th> </tr> </thead> <tbody> <tr> <td>field_analyst_agent</td> <td>领域侦测员</td> <td>自动识别论文学科，配置 3 位领域适配审稿人</td> </tr> <tr> <td>eic_agent</td> <td>主编 (Editor-in-Chief)</td> <td>综合所有意见，做出编辑决定</td> </tr> <tr> <td>methodology_reviewer_agent</td> <td>R1：方法论审稿人</td> <td>评估研究设计的严谨性</td> </tr> <tr> <td>domain_reviewer_agent</td> <td>R2：领域审稿人</td> <td>评估领域贡献和文献定位</td> </tr> <tr> <td>perspective_reviewer_agent</td> <td>R3：跨学科审稿人</td> <td>从相邻领域视角提出批判</td> </tr> <tr> <td>devils_advocate_reviewer_agent</td> <td>魔鬼代言人</td> <td>系统性挑战核心论点</td> </tr> <tr> <td>editorial_synthesizer_agent</td> <td>编辑综合者</td> <td>将 5 份审稿报告整合为统一决定</td> </tr> </tbody> </table> <hr> <h2 class="heading-element" id="field_analyst_agent自动领域侦测"><span>field_analyst_agent：自动领域侦测</span> <a href="#field_analyst_agent%e8%87%aa%e5%8a%a8%e9%a2%86%e5%9f%9f%e4%be%a6%e6%b5%8b" class="heading-mark"> <svg class="octicon octicon-link" viewBox="0 0 16 16" version="1.1" width="16" height="16" aria-hidden="true"><path d="m7.775 3.275 1.25-1.25a3.5 3.5 0 1 1 4.95 4.95l-2.5 2.5a3.5 3.5 0 0 1-4.95 0 .751.751 0 0 1 .018-1.042.751.751 0 0 1 1.042-.018 1.998 1.998 0 0 0 2.83 0l2.5-2.5a2.002 2.002 0 0 0-2.83-2.83l-1.25 1.25a.751.751 0 0 1-1.042-.018.751.751 0 0 1-.018-1.042Zm-4.69 9.64a1.998 1.998 0 0 0 2.83 0l1.25-1.25a.751.751 0 0 1 1.042.018.751.751 0 0 1 .018 1.042l-1.25 1.25a3.5 3.5 0 1 1-4.95-4.95l2.5-2.5a3.5 3.5 0 0 1 4.95 0 .751.751 0 0 1-.018 1.042.751.751 0 0 1-1.042.018 1.998 1.998 0 0 0-2.83 0l-2.5 2.5a1.998 1.998 0 0 0 0 2.83Z"></path></svg> </a> </h2><p><strong>为什么需要它</strong>：学术论文的评审标准因学科而异。一篇计量经济学论文和一篇民族志研究的质量标准完全不同。一个计算机科学 conference paper 的贡献标准和一篇医学期刊的 RCT 报告也不同。</p>https://cygnusyang.github.io/posts/01-ai-%E5%B7%A5%E5%85%B7%E4%B8%8E%E6%99%BA%E8%83%BD%E4%BD%93/08-academic-research-skills/06-reviewer/13-six-review-modes/Sat, 23 May 2026 00:00:00 +0800ruohuyang@163.com (Cygnus Yang)https://cygnusyang.github.io/posts/01-ai-%E5%B7%A5%E5%85%B7%E4%B8%8E%E6%99%BA%E8%83%BD%E4%BD%93/08-academic-research-skills/06-reviewer/13-six-review-modes/<p>Academic Paper Reviewer 提供 6 种审稿模式，从完整的多视角评审到针对性的方法论检查。其中 <code>full</code> 和 <code>methodology-focus</code> 模式在 v3.6.2 中引入了 <strong>Sprint Contract（冲刺合约）</strong> ——一种&quot;先承诺标准，再评估论文&quot;的两阶段预承诺机制，从根本上防止了审稿人&quot;看了论文后再调整标准&quot;的认知偏误。</p> <h2 class="heading-element" id="模式一览"><span>模式一览</span> <a href="#%e6%a8%a1%e5%bc%8f%e4%b8%80%e8%a7%88" class="heading-mark"> <svg class="octicon octicon-link" viewBox="0 0 16 16" version="1.1" width="16" height="16" aria-hidden="true"><path d="m7.775 3.275 1.25-1.25a3.5 3.5 0 1 1 4.95 4.95l-2.5 2.5a3.5 3.5 0 0 1-4.95 0 .751.751 0 0 1 .018-1.042.751.751 0 0 1 1.042-.018 1.998 1.998 0 0 0 2.83 0l2.5-2.5a2.002 2.002 0 0 0-2.83-2.83l-1.25 1.25a.751.751 0 0 1-1.042-.018.751.751 0 0 1-.018-1.042Zm-4.69 9.64a1.998 1.998 0 0 0 2.83 0l1.25-1.25a.751.751 0 0 1 1.042.018.751.751 0 0 1 .018 1.042l-1.25 1.25a3.5 3.5 0 1 1-4.95-4.95l2.5-2.5a3.5 3.5 0 0 1 4.95 0 .751.751 0 0 1-.018 1.042.751.751 0 0 1-1.042.018 1.998 1.998 0 0 0-2.83 0l-2.5 2.5a1.998 1.998 0 0 0 0 2.83Z"></path></svg> </a> </h2><table> <thead> <tr> <th>模式</th> <th>光谱</th> <th>审稿团队</th> <th>监管</th> <th>核心场景</th> </tr> </thead> <tbody> <tr> <td><code>full</code></td> <td>Balanced</td> <td>EIC + R1 + R2 + R3 + DA (5人)</td> <td>High</td> <td>完整同行评审</td> </tr> <tr> <td><code>methodology-focus</code></td> <td>Fidelity</td> <td>EIC + methodology (2人)</td> <td>Medium</td> <td>专注方法论审查</td> </tr> <tr> <td><code>re-review</code></td> <td>Fidelity</td> <td>EIC + synthesizer (3人精简)</td> <td>Medium</td> <td>验收修订</td> </tr> <tr> <td><code>quick</code></td> <td>Fidelity</td> <td>EIC only</td> <td>Low</td> <td>快速评估</td> </tr> <tr> <td><code>guided</code></td> <td>Originality</td> <td>EIC dialogue</td> <td>Very High</td> <td>引导式改进</td> </tr> <tr> <td><code>calibration</code></td> <td>Fidelity</td> <td>Full panel + gold set</td> <td>Medium</td> <td>评审系统校准</td> </tr> </tbody> </table> <hr> <h2 class="heading-element" id="full-mode完整多视角同行评议"><span>Full Mode：完整多视角同行评议</span> <a href="#full-mode%e5%ae%8c%e6%95%b4%e5%a4%9a%e8%a7%86%e8%a7%92%e5%90%8c%e8%a1%8c%e8%af%84%e8%ae%ae" class="heading-mark"> <svg class="octicon octicon-link" viewBox="0 0 16 16" version="1.1" width="16" height="16" aria-hidden="true"><path d="m7.775 3.275 1.25-1.25a3.5 3.5 0 1 1 4.95 4.95l-2.5 2.5a3.5 3.5 0 0 1-4.95 0 .751.751 0 0 1 .018-1.042.751.751 0 0 1 1.042-.018 1.998 1.998 0 0 0 2.83 0l2.5-2.5a2.002 2.002 0 0 0-2.83-2.83l-1.25 1.25a.751.751 0 0 1-1.042-.018.751.751 0 0 1-.018-1.042Zm-4.69 9.64a1.998 1.998 0 0 0 2.83 0l1.25-1.25a.751.751 0 0 1 1.042.018.751.751 0 0 1 .018 1.042l-1.25 1.25a3.5 3.5 0 1 1-4.95-4.95l2.5-2.5a3.5 3.5 0 0 1 4.95 0 .751.751 0 0 1-.018 1.042.751.751 0 0 1-1.042.018 1.998 1.998 0 0 0-2.83 0l-2.5 2.5a1.998 1.998 0 0 0 0 2.83Z"></path></svg> </a> </h2><p><strong>触发</strong>：&ldquo;review this paper&rdquo; / &ldquo;peer review&rdquo; / &ldquo;manuscript review&rdquo;</p>