Skip to content

joshua-zyy/academic-paper-writer / skills/academic-polishing

academic-polishing

Verified

Polish academic prose, de-AI-ify text, control claim strength, or rewrite method sections for CS/AI/ML papers. Executes Prose Quality Gate, Claim Strength Audit, and de-AI pass. Use when: removing AI writing patterns from paper text, adjusting claim strength to match evidence level, rewriting method sections with proper narrative flow, improving academic writing quality, checking for overclaiming. Triggers on: 润色, polish, improve writing, 去AI, de-AI, claim strength, 改写, rewrite method, prose quality, 降级表述, remove AI patterns, academic writing polish, 学术润色, 去AI化, 降级结论, improve prose.

references/de-ai-patterns.md

View on GitHub
# 去AI化学术文体规范

目标:让论文正文读起来像经验丰富的人类学者撰写,而不是模板驱动的机器输出。

## 空洞形容词黑名单

| 避免使用 | 条件使用 |
|---|---|
| 非常重要的、显著的 | 仅在附有数据或 VERIFIED 引用时使用;否则删除或换成具体描述 |
| 大量的、许多 | 给出具体数量或范围 |
| 各种各样的 | 列出具体类别或用 "diverse" / "heterogeneous" 等更精确的词 |
| 显而易见的、不言而喻的 | 删除或改为具体论证 |
| 至关重要的 (crucial) | 删除或改为具体机制说明 |
| 深刻的 (profound) | 删除或改为具体效应量 |
| 前所未有的 (unprecedented) | 必须有严格比较支撑,否则删除 |
| 全面的 (comprehensive) | 删除或改为具体覆盖范围 |

## AI 典型连接词替换表

| 避免使用 | 推荐替换 |
|---|---|
| 此外 (Moreover / Furthermore) | 与此相关的是、这一发现与…一致、在…方面 |
| 然而 (However) | 尽管如此、这一矛盾暗示了、与之形成对比的是 |
| 因此 (Therefore) | 这一结果表明、由此可以推断、这一现象提示了 |
| 总而言之 (In summary) | 综合以上观察、上述证据表明 |
| 值得注意的是 (It is worth noting that) | 直接陈述事实,删除套话前缀 |
| 有趣的是 (Interestingly) | 删除,直接陈述观察 |
| 重要的是 (Importantly) | 删除,用因果连接替代 |
|  delve into | 分析、探讨、检验 |
|  notably | 删除或改为具体数据支撑 |
|  significantly | 仅在附 p 值或效应量时使用 |
|  effectively | 删除或改为具体性能指标 |
|  robustly | 删除或改为具体方差/稳定性指标 |

替换原则:不要机械地把一个模板词换成另一个模板词;应根据上下文逻辑选择最自然的表述。

## Before / After 改写示例

### 示例 1:Introduction 开头
- ❌ "In recent years, deep learning has garnered significant attention across various domains."
- ✅ "Deep learning methods have achieved measurable improvements in image classification accuracy, reducing top-5 error from 15.3% to 3.6% on ImageNet between 2012 and 2015."

### 示例 2:Method 描述
- ❌ "We delve into the architecture design to effectively capture robust features."
- ✅ "To reduce sensitivity to input scale variations, we replace global pooling with adaptive spatial attention, which reweights feature maps by estimated saliency scores."

### 示例 3:Results 陈述
- ❌ "Our method significantly outperforms the baseline, demonstrating strong generalization."
- ✅ "On the held-out test set, our method achieved 87.3% accuracy (±1.2%) compared to the ResNet-50 baseline of 82.1% (±1.5%)."

### 示例 4:Discussion 段落
- ❌ "However, it is worth noting that our approach has several limitations."
- ✅ "The current evaluation relies on a single-site dataset with 312 subjects; whether the observed performance transfers to multi-site protocols remains unverified."

## 强化句子级逻辑衔接

- 不要只是罗列事实,要在句子之间建立起因果、转折或递进的深层逻辑联系。
- 常见模式:
  - 因果:前句观察 → 后句推论或解释
  - 递进:前句局限 → 后句改进或补偿
  - 转折:前句主流方法 → 后句其未解决的问题
  - 并列补足:前句方法A局限 → 后句方法B在…方面补充
- 每一段的首句应承担该段的论证职责(topic sentence),而不是空泛的背景铺垫。
- 段落末尾应自然引出下一段的论证方向,而不是突兀跳转。

## 禁用的写作口吻

正文中不得出现以下口吻:

1. **元评论口吻**:"这一节的关键是……""本文的方法学重点不在于……而在于……""下面依次解释非显然选择"
2. **审稿人对话口吻**:"该设计的合理性在于……""从当前实现看……""一个合理解释是……"
3. **代码讲解口吻**:"首先定义输入维度""然后调用 forward 函数""最后返回输出张量"
4. **checklist 痕迹**:段落读起来像是对"要写哪些要点"的逐条回答

## Prose Quality Gate 检查清单

### 通用检查

- [ ] 正文是否以完整 prose 段落为主(非提纲句、说明句)
- [ ] 是否无元评论口吻
- [ ] 是否无审稿人对话口吻
- [ ] 是否无代码讲解口吻
- [ ] 是否无空洞形容词
- [ ] AI 连接词是否已替换
- [ ] 句子间是否有深层逻辑衔接

### 章节专项

- [ ] Introduction / Related Work:有论证链和综合比较(非罗列)
- [ ] Method:有"问题→设计→机制→收益/边界"叙事(非公式罗列)
- [ ] Experimental Setup / Data:有协议、约束与风险说明
- [ ] Discussion / Limitations:有边界分析、失败模式

全部通过 → `prose_debt: closed`;任一项失败 → `prose_debt: open`

改写最多执行 2 轮;2 轮后仍未通过,保留 `prose_debt: open` 但允许继续后续步骤(最终 Verification 不得判为 passed)。
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 196kUpdated Claude CodeCodexVerifiedInstall