CUE-Bench: New Benchmark for Affective Stance in Chinese Discourse
CUE-Bench has been launched by researchers as a benchmark for Chinese Unsaid Emotion, aimed at assessing how well models grasp implicit and indirect emotional expressions in conversation. This benchmark fills a void in current emotion assessments, which mainly focus on overt sentiment or final emotional categories, neglecting the interplay between explicit expression, implicit feelings, pragmatic intent, and nuanced emotions. Centered on Affective Stance, CUE-Bench encompasses various communicative contexts, establishing nine human-readable affective stances derived from the interaction of explicit and implicit polarity, along with intent and detailed emotion annotations. The findings are available in a paper on arXiv (arXiv:2608.10810), highlighting the benchmark's goal of revealing model shortcomings in comprehending deeper emotional meanings.
Key facts
- CUE-Bench is a Chinese Unsaid Emotion benchmark focusing on Affective Stance.
- It covers diverse communicative scenarios.
- It constructs nine human-interpretable affective stances from explicit-implicit polarity interaction.
- It provides intent and fine-grained emotion annotations.
- The paper is available on arXiv with ID 2608.10810.
- The announcement type is 'cross'.
- Existing emotion benchmarks mainly annotate surface polarity or final emotion categories.
- CUE-Bench aims to address cases where affective meaning is concealed, weakened, inverted, or pragmatically reshaped.
Entities
Institutions
- arXiv