Visual Question Answering on SpuriVerse
53.79AccuracyVISTA
Evaluation Results
| Method | Links | |
|---|---|---|
| VISTAVLM=Qwen2.5-VL, Setting=RL2025.12 | 53.79 | |
| E2EVLM=Qwen2.5-VL, Setting=base + CoT2025.12 | 47.42 | |
| VISTAVLM=Llama3.2-Vision, Setting=RL2025.12 | 46.53 | |
| VISTAVLM=Qwen2.5-VL, Setting=base2025.12 | 46.29 | |
| E2EVLM=Qwen2.5-VL, Setting=RL2025.12 | 44.52 | |
| VISTAVLM=Llama3.2-Vision, Setting=base2025.12 | 44.44 | |
| E2EVLM=Llama3.2-Vision, Setting=SFT2025.12 | 40.16 | |
| E2EVLM=Llama3.2-Vision, Setting=base2025.12 | 39.76 | |
| E2EVLM=Llama3.2-Vision, Setting=base + CoT2025.12 | 38.87 | |
| E2EVLM=Qwen2.5-VL, Setting=base2025.12 | 37.5 | |
| E2EVLM=Qwen2.5-VL, Setting=SFT2025.12 | 34.84 |