Implement Average Token Log Prob with Linear Scaling #139

Closed

Assignees

Labels

Explainability Project

opened

TBD I might need to make sub-issues. Basically, this ticket confirms that we are going with the linear scaling (i.e. mapping the average token logprob to its percentile within the 4500 examples I generated).

Metadata

Assignees

taylorlynn

Labels

Explainability Project

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