Skip to content

[model-matchmakah] Learned RouteLLM-style grader #16

@holzerjm

Description

@holzerjm

Replace the self-grading confidence call with a small trained classifier per query (RouteLLM-style) and compare escalation accuracy.

Start in agentkit/route.py usage in projects/wildcards/model-matchmakah/.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions