Viewer Sentiment Prediction on Pitts (test)
40.4Clean AccuracyQwen3-VL + GRACE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen3-VL + GRACEModel Category=Fine-Tuned Open-Source MLLMs2026.06 | 40.4 | 72.7 | |
| Qwen2.5-VL + GRACEModel Category=Fine-Tuned Open-Source MLLMs, with auxiliary samples=true2026.06 | 39.6 | 73 | |
| Qwen3-VL BaselineModel Category=Fine-Tuned Open-Source MLLMs, sparse-frame fine-tuning=true2026.06 | 37.2 | 69.3 | |
| Qwen2.5-VL BaselineModel Category=Fine-Tuned Open-Source MLLMs, sparse-frame fine-tuning=true2026.06 | 35.5 | 66.7 | |
| SVMModel Category=Previous Methods and Closed-Source Models2026.06 | 32.8 | — | |
| GPT-4oModel Category=Previous Methods and Closed-Source Models2026.06 | 21.3 | 55.5 |