lexical sophistication⁠ (opens in a new tab)

Move the slider to make a language model choose more common or rarer words.

How it works

Applied linguistics measures lexical sophistication by how rare a text's words are. Word frequencies follow Zipf's law: a few words make up most of any text, and most words are rare. Here each word gets a rarity score from how often it appears in wordfreq (opens in a new tab), a large collection of English text. Common words like the score 0. Sky, seen once in 19,000 words, scores 1.3, canopy, once in 300,000, scores 2.5, and the rarest words score 4.

0.0

Qwen3.5-0.8B's likely next words. Choose one to add it.

As Qwen3.5-0.8B writes, the slider raises the chances of words rarer than an average word such as sky and lowers those of more common ones, more strongly the further a word is from average. Negative values do the opposite. When the next token only starts a word, like cob, it is judged by the words it could become.

The model runs in your browser on WebGPU, from the same graph as thought steering. Word frequencies from wordfreq by Robyn Speer, CC BY-SA 4.0.