Results for "ungrounded output"

Learning a function from input-output pairs (labeled data), optimizing performance on predicting outputs for unseen inputs.

The text (and possibly other modalities) given to an LLM to condition its output behavior.

One example included to guide output.

Explicit output constraints (format, tone).

Asking model to review and improve output.

Small prompt changes cause large output changes.

Using output to adjust future inputs.

Forcing predictable formats for downstream systems; reduces parsing errors and supports validation/guardrails.