Materials

Generate printable materials for LLMs Unplugged activities, matching the different ways of working with language models in our modules. For curated, hardbound editions of the same n-gram models, see the LLMs Unplugged Library.

Booklets (for dice generation)

Dice lookup tables for probabilistic text generation. Each entry shows a previous-words context, a dice range, and possible next words—roll the dice and read off the next word. Used in the Pre-trained Model Generation module.

The generator below handles Chinese text too, tokenised into words (or single characters, with the toggle) and ordered by pinyin—see the announcement for how the Chinese booklets work.

Worksheets

The printed grid for the Training module (grid flavour), and the training text for the Sycophancy module.

Cutouts

Printable token cards for hands-on training. Cut out the tokens, read the previous-word label on each card. Spread them on a table during the Training module (cutouts variant) — or sort them into piles as an optional extension.

Search sheets

The same matching game as the cutouts, with the cutting taken out. Each corpus is shuffled and dealt across one page per participant, so everyone searches their own sheet for the word just called out and reads out what follows it. Nobody holds the whole model — it only exists across the room — which makes these a quick way to run generation with a large group and no scissors. The Search Sheets module has the algorithm, running notes and discussion questions.

Generate your own

Upload a text file or paste text, choose your options, and download a PDF—no server required. Your text is processed entirely in your browser.

1. Enter your text: upload a file or paste text directly
2. Add metadata: provide a title (required) and optionally an author
3. Choose options: select the n-gram size and output type (booklet or cutouts)
4. Generate: preview as SVG or download as PDF

Supported file formats

• .txt — plain text files
• .md — markdown files (treated as plain text)
• .docx — Microsoft Word documents (text extracted, formatting ignored)
• .pdf — PDF documents (best-effort text extraction; scanned/image-based PDFs won’t work)