MUStReason: A Benchmark for Diagnosing Pragmatic Reasoning in Video-LMs for Multimodal Sarcasm Detection
This work addresses the challenge of pragmatic reasoning in Video-LMs for sarcasm detection, which is important for improving AI's understanding of human communication, though it is incremental as it builds on existing multimodal models.
The authors tackled the problem of multimodal sarcasm detection by introducing MUStReason, a diagnostic benchmark with annotations for modality-specific cues and reasoning steps, and proposed PragCoT, a framework that improved Video-LMs' focus on implied intentions, achieving enhanced performance in sarcasm classification.
Sarcasm is a specific type of irony which involves discerning what is said from what is meant. Detecting sarcasm depends not only on the literal content of an utterance but also on non-verbal cues such as speaker's tonality, facial expressions and conversational context. However, current multimodal models struggle with complex tasks like sarcasm detection, which require identifying relevant cues across modalities and pragmatically reasoning over them to infer the speaker's intention. To explore these limitations in VideoLMs, we introduce MUStReason, a diagnostic benchmark enriched with annotations of modality-specific relevant cues and underlying reasoning steps to identify sarcastic intent. In addition to benchmarking sarcasm classification performance in VideoLMs, using MUStReason we quantitatively and qualitatively evaluate the generated reasoning by disentangling the problem into perception and reasoning, we propose PragCoT, a framework that steers VideoLMs to focus on implied intentions over literal meaning, a property core to detecting sarcasm.