Video by Hugging Face via YouTube

The model’s top prediction can still break the rules.
I built Token Arcade to make structured output and constrained decoding easier to understand: see the next-token probabilities, pick a card, and find out whether it’s allowed.
Watch a capital letter break the “heads or tails” rule, then try building a valid JSON creature one token at a time.
Built with Transformers.js, with inference running locally in your browser through WebGPU.
Can you get 100/100?
Play: https://huggingface.co/spaces/nico-martin/tokenarcade
Code: https://github.com/nico-martin/tokenarcade