Skip to content

Picking the Best Belief Agent #7

Description

@Kenan-jeff-Jiang

I have a question regarding picking the final agent from the archive.
Currently, I set epsilon = 0.5 to have a balanced pick, so the final agent can be supported by enough evaluations and also shows promise.
But I ended up always picking the agents in the early stage of the archive. i.e., a direct child of the initial agent.
I used gpt-5.4-mini as the downstream LLM, and I set the evaluation budget to 200.

Has anyone else met the same problem, could the authors shares more insights on picking the final agent? Thanks in advance.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions