Spatial Reasoning on What'sUp (test)
95Accuracy (What'sUp Test)Gemma-3-4b-it
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gemma-3-4b-itIntervention=Ordering amp.2026.03 | 95 | 50.7 | |
| Qwen2-VL-7B-InstructIntervention=Ordering amp.2026.03 | 93 | 32.4 | |
| Gemma-3-4b-itIntervention=Random amp.2026.03 | 92 | 15.8 | |
| Gemma-3-4b-itIntervention=None2026.03 | 90 | — | |
| Qwen2-VL-7B-InstructIntervention=Random amp.2026.03 | 90 | 8.8 | |
| Qwen2-VL-7B-InstructIntervention=None2026.03 | 89 | — |