Academic Research Hub

Academic Research Hub

Overview

Academic Research Hub provides unified access to multiple academic databases (arXiv, PubMed, Semantic Scholar, etc.) for searching papers, downloading PDFs, extracting citations, and generating bibliographies. It is the recommended tool for literature review and reference management in NeuroClaw.

Prerequisite: Install OpenClawCLI (Windows, MacOS)

Installation

Best Practices:

# Standard installation
pip install arxiv scholarly pubmed-parser semanticscholar requests

# If you encounter permission errors, use a virtual environment
python -m venv venv
source venv/bin/activate   # On Windows: venv\Scripts\activate
pip install arxiv scholarly pubmed-parser semanticscholar requests

Never use --break-system-packages as it can damage your system's Python installation.

Quick Reference

Core Features

• Multi-source search (arXiv, PubMed, Semantic Scholar, Google Scholar)
• Full-text PDF download
• Citation extraction (BibTeX, RIS, JSON, plain text)
• Comprehensive metadata retrieval (title, authors, abstract, DOI, citation count, etc.)

When to Call This Skill

• Need to search for recent papers on a specific topic
• Build literature reviews or bibliographies
• Download PDFs for local reading
• Extract citations for paper-writing
• Use inside research-idea or method-design workflows

Benchmark Adapter Guidance

Use this skill as the preferred retrieval backbone for benchmark-style academic search tasks, but do not reimplement a full multi-source aggregator unless the task truly needs unsupported sources or an output schema the script does not emit directly.

• Reuse scripts/research.py first for the sources it already supports: arxiv, pubmed, and semantic.
• Treat the script outputs as normalized retrieval inputs, then add only a thin adapter layer for:
  • benchmark-specific file naming and output directories,
  • light field renaming or schema normalization,
  • per-task date-window filtering,
  • source-level merge/dedup metadata,
  • degraded-run logging when one source is unavailable.
• Do not replace the whole workflow with a brand-new all-in-one crawler if the task can be satisfied by composing this skill with a small adapter.
• If a task requires an unsupported source such as OpenReview, keep this skill for the supported sources and implement only the missing source plus the final merge/export layer.
• Preferred pattern for benchmark tasks:
  1. call research.py separately for each supported source,
  2. cache or save raw source outputs,
  3. normalize to the benchmark schema,
  4. add unsupported sources only as narrow supplemental branches,
  5. export one final task-specific JSON artifact.

Benchmark-Facing Default Mainline

For tasks like recent multi-platform paper search:

1. Use research.py arxiv ... --output ...json.
2. Use research.py pubmed ... --output ...json.
3. Use research.py semantic ... --output ...json.
4. Apply a thin adapter script to enforce:
   • last-N-days filtering,
   • newest-first sorting,
   • benchmark key names such as semantic_scholar,
   • task-level metadata and degraded status,
   • final merged export path.
5. Add OpenReview only if the task explicitly requires it and keep failure there as an expected degraded branch rather than rebuilding the entire supported-source stack.

Complementary / Related Skills

• multi-search-engine → general web/academic search

Reference

• Official documentation and APIs: arXiv, PubMed, Semantic Scholar
• OpenClawCLI: https://clawhub.ai/

Created At: 2026-03-25 18:20 HKT
Last Updated At: 2026-03-25 18:20 HKT
Author: chengwang96