Educational simulation: probabilities come from a curated browser dataset, not a live language model. Avoid entering private or sensitive text.
Game settings
Probability controls
Display and accessibility
Prediction arena
Choose the most likely continuation
Round 0/10
Score 0
Streak 0
Lives 3
Player 1: 0
Choose settings, then start the game.
How to play: choose a word, then compare its probability with the model’s top prediction. Keyboard shortcuts: 1–6 choose, H hint, N next.
Live performance
Accuracy
0%
Average time
0.0s
Average confidence
0%
Uncertainty
0.00
Prediction analysis
Custom challenge builder
Use word:probability pairs. Values are normalised automatically.
Session history
0 rounds| # | Choice | Result | Points |
|---|---|---|---|
| No rounds completed. | |||
Sentence journey
Words chosen during journey mode will appear here.
Learning guide
Conditional probability
A next-word predictor estimates which continuation fits the visible context. Each option receives a share of the total probability.
Context window
Longer context can clarify meaning, topic, grammar, and style. Shorter context often increases ambiguity and uncertainty.
Temperature
Lower temperature makes the strongest option more dominant. Higher temperature spreads probability across more creative alternatives.
Top-k and top-p
Top-k keeps a fixed number of candidates. Top-p keeps enough candidates to cover a chosen cumulative probability.
Entropy
Entropy summarises uncertainty. A low value means one option dominates, while a higher value means several words remain plausible.
Limitations
A high probability does not guarantee truth, fairness, or safety. Prediction depends on patterns and available context.