Folio AI Research Assistant
AI Research Assistant

Read papers.
Get AI summaries.
Cite sources. Organize notes.

Folio sits beside every paper you read — explaining dense sections in plain language, keeping your citations honest, and turning highlights into a connected library of notes.

No account needed for the demo · Free plan forever

Everything between “open PDF” and “finished draft”.

One quiet workspace for the whole reading loop — understand, verify, remember.

AI summaries

Plain-language explanations of any section — from abstract to appendix — grounded in the exact passage they came from.

Accurate citations

Every claim links back to its source. Copy clean APA in one click, or export your whole bibliography as BibTeX.

Organized notes

Highlights become linked notes that stay attached to their paper — searchable, taggable, and ready for your next draft.

Your reading list isn’t getting shorter.

Start free with three papers and unlimited curiosity — upgrade only when your library outgrows it.

Live demo — no account needed
Interactive paper demo

Attention Is All You Need

Vaswani, Shazeer, Parmar, et al. · NeurIPS 2017 · 62,414 citations

From the paper
× 6 ENCODER Multi-Head Self-Attention Feed Forward Embedding + Position Inputs × 6 DECODER Masked Self-Attention Cross-Attention Feed Forward Embedding + Position Outputs (shifted right) Linear + Softmax
Figure 1. The Transformer — model architecture (simplified).
AI explanation

You’re looking at the Transformer’s encoder–decoder. The left stack reads the whole input sentence at once: self-attention lets every word weigh every other word, so context never gets squeezed through a single memory the way it does in an RNN.

The right stack writes the output one token at a time, attending both to what it has written so far and to the encoder’s reading of the input (the dashed line). With no recurrence anywhere, training parallelizes across the entire sequence — the key reason Transformers train dramatically faster.

Grounded in §3 “Model Architecture” — nothing invented
Ask a question
Linked notes (preview)
Folio Already have an account? Sign in
Plans & account creation

Choose your plan

Start free. Upgrade when your library — not your curiosity — needs more room.

We’ll carry your demo workflow into your new workspace — “Attention Is All You Need” plus the notes and questions you just created.

1 · Pick a plan

2 · Create your account

Please enter your name.

That email doesn’t look right.

Use at least 8 characters.

By continuing you agree to the Terms & Privacy Policy.

My Research GRGuest Researcher

Paper Title