tools-only QUICK_REFERENCE.md — Paper Audit Tool
Quick lookup for audit modes, checks, CLI arguments, and format compatibility.
1. Check Support Matrix
| Check | .tex EN |
.tex ZH |
.typ |
.pdf |
Script |
|---|---|---|---|---|---|
format |
✅ | ✅ | ✅ | ❌ | check_format.py |
grammar |
✅ | ✅ | ✅ | ✅ | analyze_grammar.py |
logic |
✅ | ✅ | ✅ | ✅ | analyze_logic.py |
sentences |
✅ | ✅ | ✅ | ✅ | analyze_sentences.py |
deai |
✅ | ✅ | ✅ | ✅ | deai_check.py |
bib |
✅ | ✅ | ✅ | ❌ | verify_bib.py |
figures |
✅ | ✅ | ❌ | ❌ | check_figures.py |
references |
✅ | ✅ | ✅ | ❌ | check_references.py |
visual |
❌ | ❌ | ❌ | ✅ | visual_check.py |
consistency |
❌ | ✅ | ❌ | ❌ | check_consistency.py |
gbt7714 |
❌ | ✅ | ❌ | ❌ | (ZH only) |
checklist |
✅ | ✅ | ✅ | ✅ | inline Python |
Legend: ✅ = supported, ❌ = skipped/not applicable
2. Mode × Check Matrix
| Check | self-check |
review |
gate |
polish |
|---|---|---|---|---|
| format | ✅ | ✅ | ✅ | ❌ |
| grammar | ✅ | ✅ | ❌ | ❌ |
| logic | ✅ | ✅ | ❌ | ✅ |
| sentences | ✅ | ✅ | ❌ | ✅ |
| deai | ✅ | ✅ | ❌ | ❌ |
| bib | ✅ | ✅ | ✅ | ❌ |
| figures | ✅ | ✅ | ✅ | ❌ |
| references | ✅ | ✅ | ✅ | ❌ |
| visual | ✅ | ✅ | ✅ | ❌ |
| checklist | ❌ | ❌ | ✅ | ❌ |
| consistency (+ZH) | ✅ | ✅ | ✅ | ❌ |
| gbt7714 (+ZH) | ✅ | ✅ | ✅ | ❌ |
Note:
polishmode runslogic+sentencesas a fast precheck, then dispatches to Critic agent.
3. CLI Reference
Basic Usage
python audit.py <file> [options]
Positional Arguments
| Argument | Description |
|---|---|
file |
Path to document (.tex, .typ, or .pdf) |
Options
| Flag | Values | Default | Description |
|---|---|---|---|
--mode |
self-check, review, gate, polish |
self-check |
Audit mode |
--pdf-mode |
basic, enhanced |
basic |
PDF extraction engine |
--venue |
e.g. neurips, ieee, acm |
`` | Target venue for rule selection |
--lang |
en, zh |
auto-detect | Force language |
--style |
A, B, C |
A |
Polish style (A=plain, B=narrative, C=formal) |
--journal |
string | `` | Target journal for polish mode |
--skip-logic |
flag | off | Skip logic check in polish mode |
--online |
flag | off | Enable CrossRef/Semantic Scholar bib verification |
--email |
`` | Email for CrossRef polite pool | |
--scholar-eval |
flag | off | Enable ScholarEval 8-dimension assessment |
--format |
md, json |
md |
Output format (Markdown or JSON) |
--output, -o |
path | stdout | Save report to file |
Example Commands
# Quick self-check (default)
python audit.py paper.tex
# Peer review simulation targeting NeurIPS
python audit.py paper.tex --mode review --venue neurips
# Quality gate for CI/CD pipeline (JSON output)
python audit.py paper.tex --mode gate --format json --output audit.json
# Enhanced PDF audit
python audit.py paper.pdf --pdf-mode enhanced --mode self-check
# Chinese thesis with online bib verification
python audit.py thesis.tex --lang zh --online --email you@university.edu
# Polish mode precheck with style B
python audit.py paper.tex --mode polish --style B --journal "Nature"
# Full self-check with ScholarEval
python audit.py paper.tex --scholar-eval
4. Severity & Priority Reference
| Severity | Priority | Meaning | Impact on Score |
|---|---|---|---|
Critical |
P0 |
Blocking issue — must fix | -1.5 per issue |
Major |
P1 |
Significant problem | -0.75 per issue |
Minor |
P2 |
Style/convention issue | -0.25 per issue |
Score Scale (NeurIPS-style)
| Score Range | Label |
|---|---|
| ≥ 5.5 | Strong Accept |
| ≥ 4.5 | Accept |
| ≥ 3.5 | Borderline Accept |
| ≥ 2.5 | Borderline Reject |
| ≥ 1.5 | Reject |
| < 1.5 | Strong Reject |
5. Individual Script Reference
These scripts can be run standalone for targeted checks:
# Format check (strict mode)
python check_format.py paper.tex --strict
# Reference integrity (JSON output)
python check_references.py paper.tex --json
# Visual layout check
python visual_check.py paper.pdf --margin 72 --min-dpi 150
# Bibliography verification (with online lookup)
python verify_bib.py refs.bib --tex paper.tex --online --email me@uni.edu
# Sentence complexity
python analyze_sentences.py paper.tex --max-words 60 --max-clauses 3
6. PDF Extraction Modes
| Mode | Engine | Speed | Accuracy | Best For |
|---|---|---|---|---|
basic |
PyMuPDF | Fast | Good | Most papers |
enhanced |
pymupdf4llm | Slower | Better (tables/headers) | Complex layouts |
7. Output Format Reference
Markdown (default)
Human-readable report with sections for scores, issues, and checklist.
Pipe to a Markdown viewer or save with --output report.md.
JSON (--format json)
Machine-readable structured output for CI/CD integration:
{
"file": "paper.tex",
"language": "en",
"mode": "gate",
"generated_at": "2026-02-27T10:00:00",
"scores": { "quality": 5.5, "clarity": 4.75, "overall": 5.1 },
"verdict": "Accept",
"issues": [...],
"checklist": [...]
}
Last updated: 2026-02-27