Active Visual Search on EAGLE-360 1.0 (test)
64.44Accuracy (%)EAGLE-360
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| EAGLE-360Model Category=Fine-tuned2026.07 | 64.44 | 16.89 | 96.12 | 2.05 | 72.16 | 60.81 | 55.56 | 71.11 | 75 | 40 | |
| EAGLE-360Model Category=Fine-tuned, Ablation Setting=w/o GRPO2026.07 | 46.94 | 14.03 | 97.22 | 0.83 | 55.67 | 41.89 | 43.44 | 44.44 | 75 | 40 | |
| EAGLE-360Model Category=Fine-tuned, Ablation Setting=w/o RoPE Rolling2026.07 | 46.01 | 16.15 | 94.72 | 2.7 | 46.18 | 45.65 | 45.25 | 47.15 | 60 | 31.58 | |
| EAGLE-360Model Category=Fine-tuned, Ablation Setting=w/o FOV2026.07 | 39.44 | 17.54 | 94.02 | 1.11 | 48.45 | 28.38 | 41.11 | 36.67 | 50 | 40 | |
| Gemini-2.5-ProModel Category=Proprietary2026.07 | 20.28 | 48.89 | 78.61 | 18.33 | 21.65 | 4.05 | 25.56 | 25.56 | 50 | 20 | |
| HVS-3BModel Category=Fine-tuned2026.07 | 11.11 | 41.77 | 77.54 | 7.22 | 18.56 | 10.81 | 7.87 | 7.78 | 0 | 0 | |
| Qwen3-VL-4B-InstructModel Category=Open-source2026.07 | 8.33 | 54.89 | 57.22 | 6.11 | 15.46 | 1.35 | 4.44 | 11.11 | 0 | 0 | |
| Qwen3-VL-8B-InstructModel Category=Open-source2026.07 | 8.33 | 54.72 | 69.17 | 13.06 | 20.62 | 4.05 | 3.33 | 4.44 | 0 | 0 | |
| GPT-4oModel Category=Proprietary2026.07 | 7.5 | 44.26 | 80 | 6.39 | 10.31 | 2.7 | 6.67 | 8.89 | 25 | 0 | |
| InternVL3.5-8bModel Category=Open-source2026.07 | 5.28 | 82.22 | 30 | 3.89 | 19.59 | 0 | 0 | 0 | 0 | 0 | |
| InternVL3.5-4bModel Category=Open-source2026.07 | 4.72 | 77.7 | 33.06 | 4.17 | 2.06 | 2.7 | 3.33 | 11.11 | 0 | 0 | |
| Gemma-3-12b-itModel Category=Open-source2026.07 | 2.78 | 88.11 | 30.83 | 12.78 | 4.12 | 0 | 0 | 6.67 | 0 | 0 | |
| Qwen2.5-VL-7B-InstructModel Category=Open-source2026.07 | 2.78 | 101.1 | 22.6 | 20.28 | 9.28 | 0 | 1.11 | 0 | 0 | 0 | |
| Gemma-3-4b-itModel Category=Open-source2026.07 | 1.39 | 95.23 | 25 | 10.83 | 5.15 | 0 | 0 | 0 | 0 | 0 |