A team wants factual answers about current public information rather than only a model's static knowledge. Which approach best fits?

1. Evaluate a higher supported sampling temperature.
2. Set appropriate output-token limits and request the desired concise format.
3. Specify the task, relevant context, audience, and required format.
4. Evaluate grounding with Google Search for the supported model and task.