Multimodal Sentiment Analysis on MOSI (Comprehensive Metrics)
89.94Acc-2 (all)GRAMformer
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| GRAMformer2026.06 | 89.94 | 87.9 | 87.87 | 89.94 | 57.87 | 50.05 | 0.593 | 0.871 | |
| FDMA2026.06 | 88.39 | — | — | 88.36 | — | 48.52 | 0.686 | 0.822 | |
| MMML2026.06 | 88.16 | 85.91 | 85.85 | 88.15 | 56.08 | 48.25 | 0.643 | 0.838 | |
| GRAMformerEncoder Configuration=RoBERTa + Data2Vec Finetuning, Video Features=Frozen, Params=731.6M2026.06 | 87.9 | 89.94 | 89.94 | 87.86 | 57.87 | 50.05 | 0.593 | 0.0087 | |
| MMT2026.06 | 87 | 85.8 | 85.8 | 87 | — | — | 0.657 | 0.83 | |
| DEVA2026.06 | 86.29 | 84.4 | 84.48 | 86.3 | 51.78 | 46.32 | 0.73 | 0.787 | |
| MMMLEncoder Configuration=RoBERTa + Data2Vec Finetuning, Video Features=Frozen, Params=889.6M2026.06 | 85.91 | 88.16 | 88.15 | 85.85 | — | 48.25 | 0.643 | 0.0084 | |
| ConFEDE2026.06 | 85.52 | 84.17 | 84.13 | 85.52 | — | 42.27 | 0.742 | 0.784 | |
| ALMTreproduced=true2026.06 | 85.37 | 83.24 | 83.41 | 85.46 | 50.29 | 44.75 | 0.738 | 0.776 | |
| FDMER2026.06 | 84.6 | — | — | 84.7 | — | 44.1 | 0.724 | 0.788 | |
| Self-MMreproduced=true2026.06 | 84.45 | 82.54 | 82.46 | 84.44 | 52.22 | 45.56 | 0.719 | 0.794 | |
| MAG-BERT2026.06 | 84.43 | 82.37 | 82.5 | 84.61 | — | 43.62 | 0.727 | 0.781 | |
| MISAsource=obtained from [30]2026.06 | 83.54 | 81.84 | 81.82 | 83.58 | 47.08 | 41.37 | 0.776 | 0.778 | |
| MFSA2026.06 | 83.3 | — | — | 83.7 | — | 41.1 | 0.856 | 0.722 | |
| ICCN2026.06 | 83.07 | — | — | 83.02 | — | 39.01 | 0.862 | 0.714 | |
| GRAMformerEncoder Configuration=BERT Finetuning, Params=110.1M2026.06 | 82.86 | 84.54 | 84.54 | 82.8 | 50.61 | 44.17 | 0.732 | 0.7895 | |
| Self-MMEncoder Configuration=BERT Finetuning, Params=109.6M2026.06 | 82.71 | 84.48 | 84.48 | 82.66 | 53.29 | 46.18 | 0.726 | 0.7857 | |
| PMR2026.06 | 82.4 | — | — | 82.1 | — | 40.6 | — | — | |
| TETFNEncoder Configuration=BERT Finetuning, Params=110.7M2026.06 | 82.22 | 83.93 | 83.95 | 82.17 | 51.92 | 45.31 | 0.728 | 0.7896 | |
| CENetEncoder Configuration=BERT Finetuning, Params=116.0M2026.06 | 81.69 | 83.35 | 83.39 | 81.68 | 48.89 | 43 | 0.758 | 0.785 | |
| MISAEncoder Configuration=BERT Finetuning, Params=110.6M2026.06 | 81.49 | 83.02 | 83.06 | 81.46 | 47.32 | 41.46 | 0.777 | 0.775 | |
| MulTsource=obtained from [30]2026.06 | 80.95 | 79.71 | 79.63 | 80.95 | 42.68 | 36.91 | 0.879 | 0.702 | |
| LMFsource=obtained from [30]2026.06 | 79.18 | 77.9 | 77.8 | 79.15 | — | 33.82 | 0.95 | 0.651 | |
| TFNsource=obtained from [30]2026.06 | 79.08 | 77.99 | 77.95 | 79.11 | — | 34.46 | 0.947 | 0.673 | |
| MULTEncoder Configuration=Pre-extracted Features, Text Encoder=BERT, Audio/Video Features=Frozen, Params=2.57M2026.06 | 78.28 | 79.63 | 79.71 | 78.29 | 41.31 | 35.48 | 0.923 | 0.6834 | |
| GRAMformerEncoder Configuration=Pre-extracted Features, Text Encoder=BERT, Audio/Video Features=Frozen, Params=0.66M2026.06 | 78.16 | 79.36 | 79.37 | 78.12 | 38.31 | 33.5 | 0.96 | 0.6641 | |
| Graph-MFNEncoder Configuration=Pre-extracted Features, Text Encoder=BERT, Audio/Video Features=Frozen, Params=2.11M2026.06 | 78.11 | 79.51 | 79.46 | 77.99 | 36.79 | 32.95 | 0.971 | 0.6485 | |
| TFNEncoder Configuration=Pre-extracted Features, Text Encoder=BERT, Audio/Video Features=Frozen, Params=9.50M2026.06 | 77.99 | 79.21 | 79.21 | 77.92 | 39.1 | 34.2 | 0.961 | 0.6575 | |
| LMFEncoder Configuration=Pre-extracted Features, Text Encoder=BERT, Audio/Video Features=Frozen, Params=0.51M2026.06 | 77.43 | 78.66 | 78.7 | 77.4 | 38.89 | 34.72 | 0.952 | 0.6569 | |
| LF-DNNEncoder Configuration=Pre-extracted Features, Text Encoder=BERT, Audio/Video Features=Frozen, Params=0.60M2026.06 | 77.35 | 78.63 | 78.6 | 77.25 | 37.99 | 34.49 | 0.957 | 0.6566 | |
| MFNEncoder Configuration=Pre-extracted Features, Text Encoder=BERT, Audio/Video Features=Frozen, Params=2.17M2026.06 | 77.08 | 78.38 | 78.38 | 77.01 | 41.84 | 36.36 | 0.946 | 0.6576 | |
| SPT2026.06 | — | 81.2 | 81.3 | — | — | — | — | — |