Paper Quality Filter

Paper Quality Filter

Prioritize academic papers with explainable venue-level evidence. Use local references instead of relying on model memory.

Do not claim to judge the full scientific value of a paper. Rank reading priority and explain the venue evidence.

Core Principle

Rank papers by reading priority, not by absolute quality.

A matched top venue means “read this earlier.” An unmatched venue means “the local references did not confirm the venue.” It does not mean the paper is low quality.

Workflow

1. Identify the input type: pasted paper list, references, search results, CSV, JSON, Markdown, TXT, BibTeX/RIS-like text, or DOI metadata.
2. Extract paper metadata when available:

• title
• venue
• journal
• conference
• source
• container-title
• publication
• booktitle
• year
• doi
• citationCount
• abstract

1. Use venue fields first. Title-only screening is limited.
2. Run scripts/rank_papers.py for files on disk whenever possible.
3. Return papers as reading-priority tiers with evidence.
4. Add a caveat when venue metadata is missing or unmatched.

Recommendation Tiers

| Tier | 中文 | Typical Evidence | | --- | --- | --- | | Strongly Recommended | 强烈推荐 | CCF A, CAS Zone 1, JCR Q1 with IF >= 10 | | Worth Keeping | 值得保留 | CCF B, JCR Q1, CAS Zone 2, IF >= 5 | | Supplementary | 可作为补充 | CCF C, JCR Q2, EI, 北大核心, CSSCI, AMI核心, 科技核心 | | Insufficient Venue Evidence | 质量依据不足 | Missing venue, unmatched source, preprint-only source |

Always explain that Insufficient Venue Evidence / 质量依据不足 means the local reference data did not confirm the venue. It does not mean the paper is poor.

Evidence Sources

Use bundled references through script lookup:

• references/journal_scores.json: journal name, JCR quartile, impact factor, and CAS zone.
• references/ccf_conferences.json: CCF A/B/C conference ranks.
• references/eiiRankingName.json: EI-indexed source titles.
• references/chinese_journal_tags.json: Chinese core journal tags, including 北大核心, CSSCI, AMI核心, and 科技核心.

Script Usage

python scripts/rank_papers.py papers.json --lang en --format markdown --output ranked.md
python scripts/rank_papers.py papers.csv --lang zh --format json --output ranked.json
python scripts/rank_papers.py papers.bib --lang en --top 20

The script accepts .json, .csv, .md, .txt, .bib, and .ris.

JSON may be a list or an object containing one of papers, items, results, or data.

Output Shape

Start with a concise summary and coverage report, then show a table:

Title | Venue | Year | Recommendation | Evidence | Confidence

For long lists, show the top results first and offer to export the full ranked file.

Handling Uncertainty

If only titles are provided, explain that venue metadata is required for reliable screening.

If DOI-only input is provided, resolve DOI metadata when tools or internet access are available. Otherwise ask for venue, journal, or conference information.

Mark arXiv, bioRxiv, medRxiv, SSRN, and other preprint-only sources as Insufficient Venue Evidence / 质量依据不足 unless a published venue is also provided.

Do not overclaim. Venue-level metrics are screening signals, not final academic judgment.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: yilaai
• Source: yilaai/paper-quality-filter
• License: MIT

Install and usage instructions live in the source repository linked above.