Skip to content

Improve harness and reference submission #51

Description

@andreea-alexandru
  • The MRPC pairs do not reach full 128 tokens; inject some examples which do,
  • Mention leakage of the input prompt length to the server; is this acceptable?
  • Think about adding another stringent quality measure (for all classification workloads): output probability agreement up to some tolerance between decrypted version and plaintext version.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions