Skip to content

Do more evals and ablations #34

Description

@linminhtoo
  • evaluate the effect of chunk size (1024 vs 512 vs 2048)

    • on both latency & eval metrics
  • get a better sense of latency vs metric trade-offs

    • e.g. the trade-offs of changing answering effort from "normal" to "thinking"

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions