Semantic Paper Radar

Build a domain reading list from natural-language intent, then output a concise research map.

Workflow

1. Clarify query intent in one line:
• topic + scope + time window + priority (foundational / frontier / balanced).
2. Run aggregated retrieval:
• General: python3 scripts/paper_radar.py search --query "" --max 40 --years 8
• Biomedical force-on: python3 scripts/paper_radar.py search --query "" --max 40 --years 8 --biomed
3. Generate synthesis report:
• python3 scripts/paper_radar.py report --query "" --max 40 --years 8 --top 12 --mode balanced
• Biomedical force-on: add --biomed
• Export clickable HTML: add --export-html (optional --html-out )

1. Present results in Chinese unless user asked otherwise:
• 必读文献（分层）
• 学术脉络（时间线）
• 阅读顺序（先读3篇）
• 可选下一步（细分子方向）

Output Rules

• Prefer OpenAlex entries with DOI/citation metadata for "经典" judgement.
• Keep arXiv entries for "最新前沿" and unreviewed but high-momentum work.
• If the query is biomedical/clinical, explicitly include a caution that arXiv papers may be preprint.
• If retrieval is sparse, broaden query with synonyms and rerun once.

Recommended Prompt Pattern

Use this framing when user asks for recommendations:

• "按 经典奠基(3-5) + 方法跃迁(3-5) + 近两年新进展(3-5) 输出"
• "每篇给：一句贡献、为什么必读、适合第几步读"
• "最后给该领域 3 条学术脉络主线"

Optional Scholar Integration

If the environment already has a Scholar-capable tool/skill (e.g., serper-scholar), call it after report and use it only for:

• citation cross-check
• venue/author authority补充

Do not block core workflow if Scholar integration is unavailable.