Tag
This paper empirically examines when to interrupt autonomous AI agents during software execution, finding that affective-state thresholds saturate quickly, LLM judges achieve low F1 scores (0.17–0.40) at high cost, and human annotators themselves show near-chance agreement on intervention timing, making the construct unreliable as an optimization target.
This paper introduces the eJSL Dialog dataset for emotion recognition in sign language conversations, addressing the lack of conversational context in existing datasets. Benchmarking shows a domain gap when applying generic multimodal models, highlighting the need for context-aware visual extractors for sign language.
This paper presents a multimodal emotion recognition module for proactive conversational agents, using facial recognition and linguistic analysis. A user study with 20 participants reveals a 'poker face' effect where visual cues are unreliable, while linguistic analysis proves more accurate; the study also shows agents can elicit emotions through conversational adaptation.
This paper introduces GroupAffect-4, a multimodal dataset of 40 participants in 10 four-person groups performing collaborative tasks. It includes aligned physiology, eye-tracking, audio, self-report, and personality data, along with benchmark targets for within-person, between-person, and group-level analysis.
Introduces CAREBench, a benchmark grounded in appraisal theory to evaluate LLMs' emotion understanding through cognitive appraisal reasoning, revealing that current models struggle with reasoning and positive emotion recognition despite matching humans on some downstream tasks.
This paper proposes a generative framework for emotion intensity evaluation, shifting from discrete classification to continuous 0-100 scoring. It demonstrates superior performance and generalization in domains like finance.
This paper proposes a lightweight framework using sticky factorial HDP-HMMs to model conversational emotion as latent regimes from multimodal valence-arousal trajectories, aiming for interpretable and computationally efficient emotional state tracking.