LLM Lab Logo

Prompt & Controls

Tune decoding behavior and observe how the explanation changes.

Live

Temperature

0.70

Max Tokens

80

Top P

0.90

Top K

40

Rep Penalty

1.05

Result Workspace

Generated output, runtime metrics, and a beginner-friendly reflection.

No generation yet

Run a prompt to populate this workspace with output, metrics, and interpretation.

Temperature

Higher values make the sampler more willing to pick surprising tokens. Lower values make output more predictable.

Top P

Top-p keeps the smallest likely token set whose probability mass reaches the chosen threshold.

Top K

Top-k limits each choice to a fixed number of likely next tokens before sampling happens.