Repository navigation
Add results on private, pre-ChatGPT writing - #6
Merged
Merged
Conversation
… private human text Adds the local-only combination, counts of documents with a flagged passage, a table of what each detector catches when the bar is set on the certainly-human private group, and an option to keep a group's texts off the cloud step.
Counts and rates only; texts, titles and per-document scores stay outside the repository.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
Every false-positive figure so far came from public human text, which may be in the models' training data and is not the kind of writing detectors are used on.
What this adds
private_evalnow reports documents as well as passages, includes the local-only combination, can keep a group off the cloud step (--skip-group), and shows what each detector catches when the bar is set on the certainly-human private group.Results (300-word passages, bar set for 5% of public human texts)
With the bar re-set on the private passages, the classifier still catches 85% of AI text at 300 words; Clef catches 65%.
Validation
bash ci/run.sh: ruff clean, 28 tests pass.results/private_writing.mdand.jsonwere checked to contain no document ids, titles or text.Risks