New framework CLASH audits lexical vs. prosodic reliance in sarcasm detection

Researchers have developed CLASH, a new framework for evaluating spoken sarcasm detection systems. This bilingual framework uses counterfactual conditions to isolate the influence of lexical content and prosody on model predictions. Experiments with various systems, including large audio language models, indicate that lexical cues generally provide a stronger advantage for sarcasm discrimination than prosodic cues, even after duration balancing. AI

IMPACT Provides a method to better understand and potentially improve the interpretability of audio-based AI models.

RANK_REASON The cluster contains an academic paper detailing a new framework and experimental results. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• arXiv
• chlormethenmadinone acetate
• CLASH
• Hugging Face
• Mustard
• Qwen3-Omni

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →