Get Started
Home
Topics
Search
Library
Research questionHow can emotion recognition distinguish opposing reactions to the same event?Many emotion benchmarks assign one emotion to a text, obscuring cases where the same event produces positive feelings in one person and negative feelings in another. Models must therefore use the people’s differing appraisals, not just the event description, to infer their emotions.
AI
Evaluation & Benchmarks
Machine Learning
Natural Language Processing
Latest papersRecent research connected to this question, newest first.Chiaroscuro for Emotions: A Contrastive Emotion Benchmark Grounded in Appraisal TheoryThe CHIARO benchmark contains 1,000 human-annotated sentences grounded in appraisal theory, with one shared trigger eliciting opposing emotions across a ten-class taxonomy. It evaluates seven frontier LLMs and four existing emotion classifiers; the strongest LLM reaches 67.3 macro-F1, while existing classifiers perform near chance. The dataset is also tested as an added training signal for downstream emotion classification across CHIARO and ten external benchmarks.research paper · Sep 3, 2026
Related questions
How can conversational speech emotion recognition use cross-speaker context without confusing each speaker’s emotional trajectory?How can affective AI be evaluated using long-term, naturalistic, passively sensed workplace data?How can skeleton-based emotion recognition generalize to unseen performers while exposing body-motion evidence under leave-performer-out evaluation?How can real-time emotion recognition in VR handle upper-face cues occluded by head-mounted displays?