GRASP: Grounded CoT Reasoning with Dual-Stage Optimization for Multimodal Sarcasm Target Identification
This addresses the challenge of interpretable and accurate sarcasm target localization in multimodal content, which is important for applications like social media analysis, though it appears to be an incremental improvement over existing methods.
The paper tackles the problem of Multimodal Sarcasm Target Identification (MSTI), which requires precise localization of fine-grained targets in text and images, by proposing GRASP, a framework that integrates visual grounding with explicit Chain-of-Thought reasoning and dual-stage optimization, achieving state-of-the-art performance in fine-grained target identification across modalities.
Moving beyond the traditional binary classification paradigm of Multimodal Sarcasm Detection, Multimodal Sarcasm Target Identification (MSTI) presents a more formidable challenge, requiring precise localization of fine-grained targets such as textual phrases and visual regions. Existing approaches predominantly rely on implicit cross-modal alignment, offering limited interpretability and suboptimal fine-grained localization. To address these limitations, we propose GRASP, Grounded Chain-of-Thought ReAsoning with Dual-Stage Optimization for Multimodal Sarcasm Prediction and Target Identification, a framework that integrates visual grounding with explicit Chain-of-Thought (CoT) reasoning to move beyond black-box MSTI. Specifically, we curate MSTI-MAX, a refined dataset that mitigates class imbalance and enriches multimodal sarcasm cues. We introduce Grounded CoT reasoning, which explicitly anchors sarcasm-related visual regions within the reasoning trajectory and prompts the model to articulate rationales before predicting the final classification labels and sarcasm targets. Furthermore, we employ a dual-stage outcome-supervised joint optimization strategy: Supervised Fine-Tuning with a coordinate-aware weighted loss, followed by Fine-Grained Target Policy Optimization. Extensive experiments demonstrate that GRASP outperforms existing baselines in fine-grained sarcasm target identification across modalities, and an LLM-as-a-Judge evaluation quantitatively measures the quality of internal reasoning chains. Our dataset and source code will be released on GitHub.