idea-generation

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Idea Generation

研究想法生成

Generate and refine novel research ideas with literature-backed novelty assessment.

基于文献支持的新颖性评估，生成并优化创新性研究想法。

Input

输入

```
$0
```
— Research area, task description, or existing codebase context
```
$1
```
— Optional: additional context (e.g., "for NeurIPS", constraints)

```
$0
```
— 研究领域、任务描述或现有代码库上下文
```
$1
```
— 可选：额外上下文（例如，"面向NeurIPS会议"、约束条件）

Scripts

脚本

Novelty check against Semantic Scholar

基于Semantic Scholar的新颖性检查

bash

python ~/.claude/skills/idea-generation/scripts/novelty_check.py \
  --idea "Adaptive attention head pruning via gradient-guided importance" \
  --max-rounds 5

Performs iterative literature search to assess if an idea is novel.

bash

python ~/.claude/skills/idea-generation/scripts/novelty_check.py \
  --idea "Adaptive attention head pruning via gradient-guided importance" \
  --max-rounds 5

执行迭代文献搜索，评估想法是否具有新颖性。

References

参考资料

Ideation prompts (generation, reflection, novelty):

~/.claude/skills/idea-generation/references/ideation-prompts.md

构思提示（生成、反思、新颖性）：

~/.claude/skills/idea-generation/references/ideation-prompts.md

Workflow

工作流程

Step 1: Generate Ideas

步骤1：生成想法

Given a research area and optional code/paper context:

Generate 3-5 diverse research ideas
For each idea, provide: Name, Title, Experiment plan, and ratings
Use the ideation prompt templates from references

给定一个研究领域和可选的代码/论文上下文：

生成3-5个多样化研究想法
每个想法提供：名称、标题、实验计划和评分
使用参考资料中的构思提示模板

Step 2: Iterative Refinement (up to 5 rounds per idea)

步骤2：迭代优化（每个想法最多5轮）

For each idea:

Critically evaluate quality, novelty, and feasibility
Refine the idea while preserving its core spirit
Stop when converged ("I am done") or max rounds reached

对于每个想法：

批判性评估质量、新颖性和可行性
在保留核心思想的同时优化想法
当收敛（"完成"）或达到最大轮次时停止

Step 3: Novelty Assessment

步骤3：新颖性评估

For each promising idea:

Run
```
novelty_check.py
```
or manually search Semantic Scholar / arXiv
Use the novelty checking prompts from references
Multi-round search: generate queries, review results, decide
Binary decision: Novel / Not Novel with justification

对于每个有潜力的想法：

运行
```
novelty_check.py
```
或手动搜索Semantic Scholar / arXiv
使用参考资料中的新颖性检查提示
多轮搜索：生成查询词、查看结果、做出决策
二元判定：新颖/不新颖，并给出理由

Step 4: Rank and Select

步骤4：排名与选择

Score each idea on three dimensions (1-10): Interestingness, Feasibility, Novelty
Be cautious and realistic on ratings
Select the top idea(s) for development

从三个维度（1-10分）对每个想法评分：趣味性、可行性、新颖性
评分需谨慎且贴合实际
选择排名靠前的想法进行深入开发

Output Format

输出格式

json

{
  "Name": "adaptive_attention_pruning",
  "Title": "Adaptive Attention Head Pruning via Gradient-Guided Importance Scoring",
  "Experiment": "Detailed implementation plan...",
  "Interestingness": 8,
  "Feasibility": 7,
  "Novelty": 9,
  "novel": true,
  "most_similar_papers": ["paper1", "paper2"]
}