Metadata-Version: 2.4
Name: feedback-forensics
Version: 0.1.4
Summary: A tool to investigate your pairwise feedback data
Author-email: rdnfn <hi@arduin.io>, timokau <timo.kaufmann@ifi.lmu.de>
License-File: LICENSE
Classifier: License :: OSI Approved :: Apache Software License
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3
Requires-Python: >=3.10
Requires-Dist: alpaca-eval
Requires-Dist: anthropic
Requires-Dist: datasets
Requires-Dist: gradio
Requires-Dist: hydra-core
Requires-Dist: ipywidgets
Requires-Dist: jupyter
Requires-Dist: langchain
Requires-Dist: langchain-anthropic
Requires-Dist: langchain-community
Requires-Dist: langchain-openai
Requires-Dist: loguru
Requires-Dist: matplotlib
Requires-Dist: nbconvert
Requires-Dist: pandas
Requires-Dist: plotly
Requires-Dist: python-dotenv
Requires-Dist: scikit-learn
Requires-Dist: wandb
Provides-Extra: dev
Requires-Dist: black==24.10.0; extra == 'dev'
Requires-Dist: bump-my-version; extra == 'dev'
Requires-Dist: pytest>=6.0; extra == 'dev'
Description-Content-Type: text/markdown

**Feedback Forensics is a tool to investigate pairwise feedback data used for AI training and evaluation:** when used for training, what is the data teaching our models? When used for evaluation, towards what kind of models is the feedback leading us? Is this feedback asking for more lists or more ethically considerate responses? Feedback Forensics enables answering these kind of questions, building on the [Inverse Constitutional AI](https://github.com/rdnfn/icai) (ICAI) pipeline to automatically detect and measure the *implicit objectives* of annotations. Feedback Forensics is an [open-source](https://github.com/rdnfn/feedback-forensics/blob/main/LICENSE) [Gradio](https://www.gradio.app/) app that can be used both [online](https://rdnfn-feedback-forensics.hf.space) and [locally](#local-installation).

See the GitHub repo for docs and details: https://github.com/rdnfn/feedback-forensics
