SocialNLI: A Dialogue-Centric Social Inference Dataset

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Deo, Akhil, Sanders, Kate, Van Durme, Benjamin
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912632558059520
author Deo, Akhil
Sanders, Kate
Van Durme, Benjamin
author_facet Deo, Akhil
Sanders, Kate
Van Durme, Benjamin
contents Making theory-of-mind inferences from human dialogue is a strong indicator of a model's underlying social abilities, which are fundamental for adept AI assistants. However, large language and reasoning models struggle to understand sophisticated social phenomena in transcript data, such as sarcasm and irony. To assess the weaknesses of current models and to identify their solutions, we introduce SocialNLI (SoNLI) -- the first social dialogue inference dataset. SoNLI consists of a collection of dialogue transcripts hand-picked to center complex social nuances like irony and sarcasm, paired with inferences, corresponding likelihood scores, and human-written explanations. We explore social inference analysis as a facet of theory-of-mind, and evaluate LLM and reasoning model theory-of-mind ability through multi-step counterfactual reasoning.
format Preprint
id arxiv_https___arxiv_org_abs_2510_05458
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SocialNLI: A Dialogue-Centric Social Inference Dataset
Deo, Akhil
Sanders, Kate
Van Durme, Benjamin
Computation and Language
Making theory-of-mind inferences from human dialogue is a strong indicator of a model's underlying social abilities, which are fundamental for adept AI assistants. However, large language and reasoning models struggle to understand sophisticated social phenomena in transcript data, such as sarcasm and irony. To assess the weaknesses of current models and to identify their solutions, we introduce SocialNLI (SoNLI) -- the first social dialogue inference dataset. SoNLI consists of a collection of dialogue transcripts hand-picked to center complex social nuances like irony and sarcasm, paired with inferences, corresponding likelihood scores, and human-written explanations. We explore social inference analysis as a facet of theory-of-mind, and evaluate LLM and reasoning model theory-of-mind ability through multi-step counterfactual reasoning.
title SocialNLI: A Dialogue-Centric Social Inference Dataset
topic Computation and Language
url https://arxiv.org/abs/2510.05458