research-tracking

name	research-tracking
description	Research direction tracking, trend discovery, and author monitoring.
metadata	{"nanobot": {"always": false}}

Research Tracking Skill

This skill helps you monitor research directions, discover emerging trends, and track influential authors.

When to Use This Skill

Activate when the user asks for:

• "最近 XXX 有什么新进展？"
• "帮我追踪一下 XXX 方向"
• "XXX 领域现在最火的是什么？"
• "XXX 团队最近在做什么？"
• "有没有什么跨学科的机会？"

Trend Discovery Methods

Method 1: Keyword Frequency Analysis

Track which keywords are appearing more frequently in recent papers:

1. Search for papers in the last 6 months on arXiv
2. Extract key terms from titles and abstracts
3. Compare with 1-2 years ago to identify rising terms

Example: If "speculative decoding" appears in 5% of LLM inference papers in 2024 vs. 0% in 2023, it's an emerging trend.

Method 2: Citation Velocity

Papers gaining citations rapidly (relative to their age) indicate hot topics:

• Use Semantic Scholar API to get citation counts and velocity
• Compare across subfields

Method 3: Conference Proceedings Scan

Top conference acceptance patterns reveal trends:

• NeurIPS/ICML/ICLR: What topics have dedicated workshops?
• ACL/EMNLP: What shared tasks are new this year?
• CVPR/ICCV/ECCV: What challenges are being organized?

Method 4: Cross-Pollination Detection

Breakthroughs often come from borrowing ideas across fields:

• Look for papers combining previously separate areas
• Example: "Transformer + graph neural networks" → Graph Transformers
• Example: "Diffusion models + protein design" → new bio-AI intersection

Author Tracking Strategy

Identifying Key Authors

1. Pioneers: First paper on the topic, highly cited
2. Prolific contributors: 3+ papers on the topic in recent years
3. Influential critics: Papers that spark debate or correction

Tracking Workflow

1. Use academic_search with au:AuthorName to find their recent papers
2. Use get_related_papers on their most influential work
3. Check their collaborators (co-authors) for related work

Author Alert Setup

Guide the user to set up research_feed in config:

{
  "research": {
    "research_feed": {
      "enabled": true,
      "feeds": [
        {
          "source": "arxiv",
          "categories": ["cs.CL", "cs.LG"],
          "keywords": ["your keyword 1", "your keyword 2"],
          "schedule": "0 8 * * *",
          "max_results": 5,
          "notify_channel": "feishu"
        }
      ]
    }
  }
}

Trend Report Format

When reporting trends to the user, use this structure:

# 🔥 [Topic] 最新趋势报告

## 热门方向
1. **[Trend Name]** — [Brief description]
   - 代表论文: [Paper 1], [Paper 2]
   - 热度指标: [Citations, frequency, etc.]

2. **[Trend Name]** — ...

## 值得关注的论文
| 论文 | 作者 | 核心创新 | 为什么重要 |
|------|------|---------|-----------|
| ...  | ...  | ...     | ...       |

## 新兴交叉领域
- **[Intersection A + B]** — [Description and potential]

## 建议行动
1. [具体建议：阅读某篇论文、关注某个作者、尝试某个方法]

Research Profile Maintenance

If memory/research_profile.md exists:

• Update it with newly discovered keywords and authors
• Mark topics that are becoming less relevant
• Add new research questions that emerge from trend analysis

Proactive Tracking Tasks

You can suggest adding these to HEARTBEAT.md:

• Weekly arXiv scan for top keywords
• Monthly citation velocity check for tracked papers
• Quarterly trend summary report