AI论文写作与投稿实战¶

Table 1: Comparison with state-of-the-art methods on [Dataset].
† indicates results reproduced by us. Best in **bold**, second best _underlined_.

| Method      | Venue    | Backbone  | Metric1 | Metric2 | Metric3 |
|-------------|----------|-----------|---------|---------|---------|
| MethodA [1] | CVPR'24  | ResNet-50 | 85.2    | 82.1    | 83.6    |
| MethodB [2] | NeurIPS'24| ViT-B    | 87.5    | 84.3    | 85.9    |
| MethodC [3] | ICLR'25  | ViT-B    | _89.1_  | _86.7_  | _87.9_  |
| **Ours**    | -        | ViT-B    | **91.3**| **88.5**| **89.9**|

Table 2: Ablation study on [Dataset]. Each row removes one component.

| Setting          | Metric1 | Metric2 | Δ Metric1 |
|------------------|---------|---------|-----------|
| Full Model (Ours)| **91.3**| **88.5**| -         |
| w/o Module A     | 89.7    | 87.1    | -1.6      |
| w/o Module B     | 90.1    | 87.8    | -1.2      |
| w/o Module A & B | 87.5    | 85.2    | -3.8      |
| replace A → naive| 88.9    | 86.5    | -2.4      |

**应对策略**：明确区分，用表格对比

回应模板：
"We appreciate R[X]'s comment and clarify the key differences between
our method and [XX]:

| Aspect       | [XX]              | Ours               |
|--------------|-------------------|--------------------|
| 目标问题      | [问题A]           | [问题B]             |
| 核心机制      | [机制A]           | [机制B]             |
| 适用场景      | [场景A]           | [场景B]             |
| 计算复杂度    | O(n²)            | O(n log n)          |

The fundamental difference lies in [本质区别]. While [XX] focuses on
[方向A], our work addresses [方向B] which requires [不同的技术考量]."

**应对策略**：补充实验 + 解释选择理由

回应模板：
"Thank you for this suggestion. We have conducted additional experiments:

(1) New dataset: [Dataset X]
| Method | Acc. | F1 |
|--------|------|----|
| Baseline | 85.2 | 83.1 |
| Ours | **88.7** | **86.5** |

(2) New baseline: [Method Y] (published at [Venue'2025])
Our method outperforms [Method Y] by [X]% on [Metric].

We chose [original datasets] because they are the most widely-used
benchmarks in [领域], used by [列举近期论文]. We will include the
above results in the camera-ready version."

**应对策略**：补充消融 + 解释每个组件的贡献

回应模板：
"We have added the requested ablation study:

| Variant | Metric1 | Metric2 | 说明 |
|---------|---------|---------|------|
| Full    | **91.3**| **88.5**| 完整模型 |
| w/o A   | 89.7    | 87.1    | 去掉模块A，性能下降1.6% |
| w/o B   | 90.1    | 87.8    | 去掉模块B，性能下降1.2% |
| A→naive | 88.9    | 86.5    | 替换为简单版本，下降2.4% |

This demonstrates that [模块A] contributes [X]% improvement
primarily through [机制], while [模块B] provides [Y]% gain via [机制]."

**应对策略**：多角度论证提升的意义

回应模板：
"We respectfully note that:

(1) **统计显著性**：We run 5 seeds and report mean±std. The improvement
    is statistically significant (p<0.01 via paired t-test).

(2) **历史对比**：Recent top-venue papers in [领域] report similar
    margins of improvement:
    - [Paper A, ICLR'25]: +0.8% over previous SOTA
    - [Paper B, NeurIPS'25]: +1.2% over previous SOTA
    - Ours: +[X]% over previous SOTA

(3) **效率优势**：Beyond accuracy, our method achieves [X]× speedup
    with [Y]% fewer parameters, making it more practical.

(4) **困难场景**：On the challenging subset [XX], our improvement
    is more pronounced (+[Z]%)."

**应对策略**：补充不同规模/领域的实验

回应模板：
"We have evaluated scalability across different settings:

| Model Scale | Params | Dataset A | Dataset B | Dataset C |
|-------------|--------|-----------|-----------|-----------|
| Small       | 12M    | 87.2      | 84.5      | 82.1      |
| Base        | 86M    | 91.3      | 88.5      | 86.7      |
| Large       | 307M   | 93.1      | 90.8      | 89.2      |

Our method consistently improves across all scales, demonstrating
strong scalability. We also tested on out-of-domain [Dataset D]
(0-shot): [结果], showing promising generalization."

**应对策略**：承认concurrent work，强调独立性和互补性

回应模板：
"Thank you for pointing out this concurrent work. We note that:

(1) Our submission (May 2025) and [concurrent work] (arXiv June 2025)
    were developed independently and concurrently.
(2) Key differences:
    - We focus on [方面A], while they focus on [方面B]
    - Our [技术X] is distinct from their [技术Y]
(3) We have added a comparison in the related work section and will
    include experimental comparison in the camera-ready version."

## 投稿前最终检查

### 内容检查
- [ ] Abstract控制在字数限制内
- [ ] 所有Contribution在实验中都有对应验证
- [ ] Ablation覆盖了所有核心模块
- [ ] Related Work没有遗漏重要baseline
- [ ] 结论不超出实验支撑的范围

### 格式检查
- [ ] 严格遵循会议模板（页数、字体、边距）
- [ ] 图片清晰度≥300dpi
- [ ] 表格使用booktabs三线表
- [ ] 参考文献格式统一、信息完整
- [ ] 正文中引用了每一个图/表

### 匿名检查（双盲审稿）
- [ ] 论文中无作者姓名和单位
- [ ] 代码链接已匿名化（Anonymous GitHub）
- [ ] 自引使用第三人称（"The work of [X]" 而非 "Our prior work [X]"）
- [ ] 元数据中无作者信息
- [ ] 致谢部分隐藏

### 技术检查
- [ ] 所有公式编号正确
- [ ] 超参数设置完整记录
- [ ] 随机种子固定，实验可复现
- [ ] Supplementary material准备完毕