Research map / Vision-language and multimodal / Cross-modal learning
ConsMSA: Semantic Distribution Consistency Learning for Multimodal Sentiment Analysis
Pan Wang, Lipeng Ke, Huajun Ying, Pritish Mohapatra, Rohan Sarkar, Suresh Lakhani, sankar venkataraman, Jingtong Hu
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 |
|---|---|---|---|
| add IntraCS term (ablation) | ConsMSA · CMU-MOSI (T+A, RoBERTa) | Acc-7: 47.52 -> 51.46 | Improved |
| add InterCS term (ablation) | ConsMSA · CMU-MOSI (T+A, RoBERTa) | Acc-7: 46.79 -> 51.46 | Improved |
| add soft gating (ablation) | ConsMSA · CMU-MOSI (T+A, RoBERTa) | Acc-7: 45.77 -> 51.46 | Improved |
| add relevance term Rel in loss/signal (ablation) | ConsMSA · CMU-MOSI (T+A, RoBERTa) | Acc-7: 45.48 -> 51.46 | Improved |
| add modality-specific loss (ablation) | ConsMSA · CMU-MOSI (T+A, RoBERTa) | Acc-7: 47.38 -> 51.46 | Improved |
| add I2CS consistency loss (ablation) | ConsMSA · CMU-MOSI (T+A, RoBERTa) | Acc-2: 86.59 -> 90.09 | Improved |
| use JS-based I2CS signal instead of entropy/attention/random masking (ablation) | ConsMSA · CMU-MOSI (T+A, RoBERTa) | Acc-7: 49.27 (random) -> 51.46 | Improved |
| use cosine or L2 distance for masking (ablation) | ConsMSA · CMU-MOSI (T+A, RoBERTa) | Acc-7: 47.96 (Cosine), 47.23 (L2) vs 51.46 | Hurt |
| hard selection retaining 80% tokens vs soft (1.0) (ablation) | ConsMSA · CMU-MOSI | Acc-2: 90.09 -> 90.85 | Improved |
| hard selection retaining 80% tokens vs soft (1.0) (ablation) | ConsMSA · CH-SIMS | Acc-2: 83.59 -> 82.06 | Hurt |
| prune to 10% of tokens (ablation) | ConsMSA · CMU-MOSI | Acc-2: 90.09 -> 89.18 | No clear effect |
| prune to 10% tokens | ConsMSA (RoBERTa) · MOSI latency | speedup: 2.16x (42.84 ms) | Improved |
Closest papers by meaning
- SemMSA: Latent Semantic-Aided Robust Multimodal Sentiment Analysis with Incomplete Data (NeurIPS 2026 · similarity 0.91)
- Towards Multimodal Sentiment Analysis via Hierarchical Correlation Modeling with Semantic Distribution Constraints (AAAI 2025 · similarity 0.90)
- A Unified Self-Distillation Framework for Multimodal Sentiment Analysis with Uncertain Missing Modalities (AAAI 2024 · similarity 0.89)
- Toward Robust Incomplete Multimodal Sentiment Analysis via Hierarchical Representation Learning (NeurIPS 2024 · similarity 0.89)
- DLF: Disentangled-Language-Focused Multimodal Sentiment Analysis (AAAI 2025 · similarity 0.89)
- Towards Robust Multimodal Sentiment Analysis with Incomplete Data (NeurIPS 2024 · similarity 0.88)
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.
