Skip to content

Latest commit

 

History

History
8 lines (6 loc) · 499 Bytes

README.md

File metadata and controls

8 lines (6 loc) · 499 Bytes

M-HalDetect

A dataset used for detecting and preventing hallucinations in multimodal models. Original paper

Each sample contains the exact question asked to the multimodal model, with the coco 2014 validation image id that was used. The annotations for each sample consists of a list of segments, containing start and end indices, with a string representing the spanned text, and a class label.

Terms of Use

See the license for terms of use.