A Faithful Benchmark for Information-Seeking Dialogue
- Type
- dataset
- Venue
- McGill-NLP
- Year
- 2026
- Source
- huggingface
- Access
- free
- Language
- en
- Added
- 2026-07-17T19:58:41.419207+00:00
- Verified
- 2026-07-17T19:58:41.419207+00:00
Summary
FaithDial is a faithful knowledge-grounded dialogue benchmark, composed of **50,761** turns spanning **5649** conversations. It was curated through Amazon Mechanical Turk by asking annotators to amend hallucinated utterances in [Wizard of Wikipedia](https://parl.ai/projects/wizard_of_wikipedia/) (WoW). In our dialogue setting, we simulate interactions between two speakers: **an information seeker** and **a bot wizard**.
Keywords
hf-dataset chat---dialogue · -language-modeling dialogue-modeling crowdsourced monolingual text datasets mlcroissant faithful-dialogue-modeling trustworthy-dialogue-modeling has-paper
Topics
Chat / Dialogue, Language Modeling
Research notes
- downloads=364; likes=18