Visual Search and Reasoning on VStar
76.96ScoreOurs-NonMath
Evaluation Results
| Method | Links | |
|---|---|---|
| Ours-NonMathBudget=10%, Model=Qwen3-VL-2B2026.03 | 76.96 | |
| Ours-MixedBudget=10%, Model=Qwen3-VL-2B2026.03 | 76.96 | |
| Ours-MathBudget=10%, Model=Qwen3-VL-2B2026.03 | 74.35 | |
| Ours-MathBudget=25%, Model=Qwen3-VL-2B2026.03 | 69.63 | |
| Ours-MixedBudget=25%, Model=Qwen3-VL-2B2026.03 | 69.63 | |
| InterlaceBudget=10%, Model=Qwen3-VL-2B2026.03 | 68.59 | |
| RandomBudget=10%, Model=Qwen3-VL-2B2026.03 | 64.4 | |
| Ours-NonMathBudget=25%, Model=Qwen3-VL-2B2026.03 | 63.87 | |
| CKABudget=10%, Model=Qwen3-VL-2B2026.03 | 62.83 | |
| RandomBudget=25%, Model=Qwen3-VL-2B2026.03 | 58.12 | |
| Ours-NonMathBudget=40%, Model=Qwen3-VL-2B2026.03 | 54.97 | |
| InterlaceBudget=25%, Model=Qwen3-VL-2B2026.03 | 50.79 | |
| CKABudget=25%, Model=Qwen3-VL-2B2026.03 | 49.74 | |
| InterlaceBudget=40%, Model=Qwen3-VL-2B2026.03 | 46.6 | |
| Ours-MathBudget=40%, Model=Qwen3-VL-2B2026.03 | 39.79 | |
| CKABudget=40%, Model=Qwen3-VL-2B2026.03 | 39.27 | |
| Ours-MixedBudget=40%, Model=Qwen3-VL-2B2026.03 | 37.17 | |
| RandomBudget=40%, Model=Qwen3-VL-2B2026.03 | 35.08 |