New COLM 2026 Paper Releases 76K Bilingual Multi-Party Dialogue Dataset

imjuhokim · x · 2026-08-05

Researchers have introduced TIDES, a new in-the-wild, longitudinal, multi-party collaboration dialogue dataset. It contains roughly 76,000 utterances in both English and Korean. The dataset is designed to help model multi-party social dynamics and is detailed in a paper published at COLM 2026.

Original post →

More from Research

Research channel →