Deep notes, not recap
The note tracks what the paper proves, what it does not prove, which experiments matter, and where the conclusions are bounded.
Turn one dense paper into an Obsidian note worth keeping.
DeepPaperNote is an evidence-first reading workflow for researchers who want more than an abstract summary.
The output is the product.
DeepPaperNote is built around the note you keep: structured, evidence-aware, figure-conscious, and ready for a long-term Obsidian vault.
Clear result tables
Core comparisons across models, datasets, tasks, settings, or metrics become compact Markdown tables with interpretation after the numbers.
Image-first figures
Usable figure and table candidates are embedded as real images; placeholders stay for missing, broken, or mismatched visuals.
Generated structure
DeepPaperNote output
- 01 Core metadata
- 02 Evidence-backed claims
- 03 Compact result table
- 04 Image-first figures
- 05 Deep analysis
Results, figures, and formulas
Comparisons, visuals, and math stay close to the analysis.
Core comparisons become compact Markdown tables, while usable visuals and formulas stay near the claims they support.
Fig. 2 Architecture + Table 4 Results image inserted
Input tokens
Embedding
Multi-head attention residual + norm
Feed-forward
Output states
Attention(Q, K, V) = softmax(QKT / √dk)V
| Setting | Metric | What it means |
|---|---|---|
| Main model | +4.8 F1 | Best overall trade-off |
| Ablation | -2.1 F1 | Module matters |
| Baseline | 73.2% | Reference point |
A disciplined path from paper to note.
Scripts prepare structured evidence. The model does the reading. The final note is linted, reviewed, and saved where your research actually lives.
- Resolve Pick one canonical paper identity.
- Source Read from source sections and metadata, not just title or abstract.
- Plan Decide the central claims, key experiments, and analysis focus.
- Results & figures Turn comparisons into tables and embed usable visuals.
- Review Check grounding, analytical depth, and readability before saving.
- Save Write to an Obsidian-style paper folder.
Install the skill. Hand it one paper.
A title, DOI, URL, arXiv link, or local PDF is enough to start a deep-reading note.
npx skills add 917Dhj/DeepPaperNote python3 -m pip install PyMuPDF Turn this paper into an Obsidian note: https://arxiv.org/abs/1706.03762