←
SIGdial293 posts
SIGdial avatar

SIGdial

@sigdial

The 27th Annual Meeting of the Special Interest Group on Discourse and Dialogue, 3-5 August 2026. Atlanta, Georgia, USA. #SIGDIAL2026

◷ Joined April 2017▧ 82 Photos and videos
SIGdial avatar
SIGdial@sigdial·Nov 6

SIGdial brings together discourse and dialogue researchers from NLP and speech. It runs the SIGDIAL conference yearly and YRRSDS for young researchers, and sponsors other events. Members get event news, job openings and a vote for the board. Learn more: sigdial.org/membership/

SIGdial avatar
SIGdial@sigdial·Oct 20

Better Scores, Worse Grounding: fine-tuning made dialogue fact-checkers score higher overall while quietly breaking cases where a pronoun points back to an earlier turn. Worth looking past the headline number. Park & Zubiaga aclanthology.org/2026.sigdia…

SIGdial avatar
SIGdial@sigdial·Oct 20

A voice agent that interviews older adults before a dementia evaluation. Tested with 30 participants from a neurology clinic, it picked up symptoms well and people liked using it. Breithaupt et al. aclanthology.org/2026.sigdia…

SIGdial avatar
SIGdial@sigdial·Oct 20

Captions tell blind users what's in a meme but tend to kill the joke. MemeBuddy turns memes into a short back-and-forth between two voices, and blind participants found it more engaging. Bhansali, Ashok & Lee aclanthology.org/2026.sigdia…

SIGdial avatar
SIGdial@sigdial·Oct 20

Seeing the same thing isn't the same as understanding each other. Some vision-language models assume dialogue partners are on the same page just because of what's on the map. Li, Gatt & Poesio aclanthology.org/2026.sigdia…

SIGdial avatar
SIGdial@sigdial·Oct 20

Too Polite to Disagree: when LLM agents work on a problem together, the ones that cave easily can drag the group toward wrong answers. Telling each agent how sycophantic its peers are raised accuracy 10.5 points. Kasprova et al. aclanthology.org/2026.sigdia…

SIGdial avatar
SIGdial@sigdial·Oct 6

When a model predicts who speaks next, is there only one right answer? Kubo et al. argue turn-taking is inherently ambiguous and propose scoring models' predictions as distributions, not yes/no labels. Check the paper here: aclanthology.org/2026.sigdia… #SIGDIAL2026 #AI #NLP

SIGdial avatar
SIGdial@sigdial·Apr 4

⏰ Deadlines approaching for #SIGDIAL2026 Dialog researchers: only 9 days until abstracts are due! 📅 Key dates (AoE): • Apr 13 - Title & abstract • Apr 20 - Full paper (PDF) • May 25 - ARR commitment • Jun 1 - Notifications 📍 Atlanta, GA | Aug 2–5 @ Emory University

Post attachment
SIGdial avatar
SIGdial@sigdial·Feb 25

Date Change due to the FIFA World Cup! 📢 SIGdial will now be held on the 3rd to 5th of August 🎉 Deadlines are updated accordingly: • Apr 13 - Title & abstract • Apr 20 - Full paper (PDF) • Mav 25 - ARR commitment • Jun 15- Notifications Workshops on the 1st and 2nd Aug

Post attachment
SIGdial avatar
SIGdial@sigdial·Feb 1

📢 Call for Papers: #SIGDIAL2026 Submit your latest work on discourse & dialogue to the 27th Annual SIGdial Conference (Atlanta, July 13–15, 2026)! 🗓 Deadlines (AoE) • Mar 23 – Title & abstract • Mar 30 – Full paper (PDF) • May 4 – ARR commitment • May 25 – Notifications

Post attachment