Research map / Vision-language and multimodal / Cross-modal learning
Conditional Information Bottleneck for Multimodal Fusion: Overcoming Shortcut Learning in Sarcasm Detection
Yihua Wang, Qi Jia, Cong Xu, Feiyu Chen, Yuhan Liu, Haotian Zhang, Liang Jin, Lu Liu, Zhichun Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| remove shortcut cues (MUStARD++R) | SpeechPrompt v2 · MUStARD++ vs MUStARD++R | weighted F1: 73.47 -> 43.59 | Hurt |
| remove shortcut cues | ITHP · MUStARD++ vs MUStARD++R | weighted F1: 70.93 -> 68.27 | Hurt |
| remove shortcut cues | TBJE · MUStARD++ vs MUStARD++R | weighted F1: 71.82 -> 68.75 | Hurt |
| remove shortcut cues | SIB · MUStARD++ vs MUStARD++R | weighted F1: 69.39 -> 68.21 | Hurt |
| remove shortcut cues | DIB · MUStARD++ vs MUStARD++R | weighted F1: 70.98 -> 70.42 | No clear effect |
| remove shortcut cues | GPT-4o · MUStARD++ vs MUStARD++R | weighted F1: 71.74 -> 75.19 | Improved |
| remove shortcut cues | Gemini 2.5 · MUStARD++ vs MUStARD++R | weighted F1: 72.20 -> 73.73 | Improved |
| remove shortcut cues | MCIB · MUStARD++ vs MUStARD++R | weighted F1: 76.85 -> 75.64 | No clear effect |
| use transformer encoder instead of MLP (ablation) | MCIB · MUStARD++R | weighted F1: 74.32 -> 75.64 | Improved |
| use fine-grained word-level features instead of coarse-grained (ablation) | MCIB · MUStARD++R | weighted F1: 71.19 -> 75.64 | Improved |
| add auxiliary modality via CIB (text -> audio+text) (ablation) | MCIB · MUStARD++R | weighted F1: 70.98 -> 73.69 | Improved |
| use pairing va + at + tv instead of vt + av + ta (ablation) | MCIB · MUStARD++R | weighted F1: 74.03 -> 75.64 | Improved |
Closest papers by meaning
- MMSD3.0: A Multi-Image Benchmark for Real-World Multimodal Sarcasm Detection (CVPR 2026 · similarity 0.90)
- Debiasing Multimodal Sarcasm Detection with Contrastive Learning (AAAI 2024 · similarity 0.89)
- G^2SAM: Graph-Based Global Semantic Awareness Method for Multimodal Sarcasm Detection (AAAI 2024 · similarity 0.88)
- DocMSU: A Comprehensive Benchmark for Document-Level Multimodal Sarcasm Understanding (AAAI 2024 · similarity 0.87)
- Well, Now We Know! Unveiling Sarcasm: Initiating and Exploring Multimodal Conversations with Reasoning (AAAI 2024 · similarity 0.86)
- A Unified Self-Distillation Framework for Multimodal Sentiment Analysis with Uncertain Missing Modalities (AAAI 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
