license: odc-by
pretty_name: FF Public App Results
extra_gated_prompt: >-
**Note: This repo is no longer in use since Feedback Forensics v0.4.5.** The
latest public results are available (ungated) at
[rdnfn/ff-public-results](https://huggingface.co/datasets/rdnfn/ff-public-results)
and can be installed with `feedback-forensics --load-web-datasets` (v0.4.5 or
higher, no token required). Consider upgrading your Feedback Forensics
installation to download this public data. If you're looking for the dataset
contribution discussed in our (upcoming) paper, this dataset can be found
here:
[rdnfn/feedback-forensics-annotations](https://huggingface.co/datasets/rdnfn/feedback-forensics-annotations).
Reach out via [arduin.io](arduin.io) if you are looking for a different
dataset or think access would be useful.
Feedback Forensics Public App Results
Note: This repo is no longer in use since Feedback Forensics v0.4.5. The latest public results are available at rdnfn/ff-public-results and can be installed with feedback-forensics --load-web-datasets (v0.4.5 or higher). Consider upgrading your Feedback Forensics installation to download this public data.
This repo contains the datasets shown publicly in the Feedback Forensics web app. This dataset was created for serving data to the public web app, rather than sharing the data itself more broadly. Due licensing and the partially safety-related nature of the data, access is gated at the moment.
If you're looking for the dataset contribution discussed in our (upcoming) paper, this dataset can be found here: rdnfn/feedback-forensics-annotations.
The annotations are provided over pairwise model outputs, typically consisting of a prompt and two responses.
Each individual annotation considers as single personality trait (e.g. "confidence"). The annotation indicates when
two responses differ with respect to that trait
(e.g. one response is more confident), a trait does not apply to both responses or is equal in both responses.
Annotations were collected using the Inverse Constitutional AI (ICAI) annotation pipeline with a
fixed set of personality traits to be tested, using gpt-4o-mini-2024-07-18. The data is provided in the AnnotatedPairs json data
format, described here.
License
This dataset is licensed under Open Data Commons License Attribution License (ODC-By) (https://opendatacommons.org/licenses/by/1-0/), unless other licenses apply.