Dense Object Detection on Dense200
85.5F1 Score @ IoU 0.5VisHarness (Qwen3-VL-8B)
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| VisHarness (Qwen3-VL-8B)Type=Agent2026.05 | 85.5 | 14.7 | — | 69.3 | — | |
| VisHarness-SFT (Qwen3-VL-8B)Type=Agent2026.05 | 83.3 | 14.5 | — | 67.9 | — | |
| Rex-Omni-3B2026.05 | 78.4 | 10.3 | 58.3 | — | — | |
| Rex-OmniType=Specialist2026.05 | 78.4 | 10.3 | — | 58.3 | — | |
| SEED1.5-VL2026.05 | 76.9 | 5.3 | 53.2 | — | — | |
| SEED1.5-VLType=MLLM2026.05 | 76.9 | 5.3 | — | 53.2 | — | |
| LocateAnything-3B2026.05 | 74 | 18.5 | 58.7 | — | — | |
| Rex-Omni-SFT-3B2026.05 | 60.2 | 10.6 | 46.4 | — | — | |
| Rex-Omni-SFTType=Specialist2026.05 | 60.2 | 10.6 | — | 46.4 | — | |
| Grounding DINO-Swin-TScore Thresh.=0.252026.05 | 36.9 | 19.7 | 33.1 | — | — | |
| Grounding DINO-Swin-TType=Open-set, Backbone=Swin-T2026.05 | 36.9 | 19.7 | — | 33.1 | — | |
| MiMo-VL-7B2026.05 | 29.7 | 0.4 | 15.9 | — | — | |
| MiMo-VL-7BType=MLLM2026.05 | 29.7 | 0.4 | — | 15.9 | — | |
| Cosmos-Reason2-8B2026.05 | 25.1 | 1.1 | 15.1 | — | — | |
| OVIS2.5-2B2026.05 | 17.9 | 0 | 6.7 | — | — | |
| OVIS2.5-2BType=MLLM2026.05 | 17.9 | 0 | — | 6.7 | — | |
| Qwen3-VL-4B2026.05 | 17.5 | 2.4 | 12.5 | — | — | |
| Qwen3-VL-8BType=Agent, with tools=true2026.05 | 16.5 | 1.5 | — | 12.9 | — | |
| DeepSeek-VL2-Small2026.05 | 16 | 3.9 | 12.7 | — | — | |
| DeepSeek-VL2-SmallType=MLLM2026.05 | 16 | 3.9 | — | 12.7 | — | |
| OVIS2.5-9BType=MLLM2026.05 | 14 | 0 | — | 5.1 | — | |
| Qwen3-VL-8B2026.05 | 13.5 | 1.7 | 9.6 | — | — | |
| DeepSeek-VL2-TinyType=MLLM2026.05 | 2.2 | 0.3 | — | 1.5 | — | |
| Qwen2.5-VL-7BType=MLLM2026.05 | 1.1 | 0.1 | — | 0.6 | — | |
| Qwen2.5-VL-3BType=MLLM2026.05 | 0.8 | 0.1 | — | 0.5 | — | |
| BagelOutput Format=bbox2026.07 | — | — | — | — | 42.4 | |
| Grounding DINO-Swin-TOutput Format=bbox2026.07 | — | — | — | — | 33.1 | |
| LocateAnythingOutput Format=bbox2026.07 | — | — | — | — | 58.7 | |
| Qwen3-VL-8B-InstructOutput Format=bbox2026.07 | — | — | — | — | 13.5 | |
| Qwen3.5-9BOutput Format=bbox2026.07 | — | — | — | — | 27.5 | |
| Rex-OmniOutput Format=bbox2026.07 | — | — | — | — | 58.3 | |
| SenseNova-VisionOutput Format=bbox2026.07 | — | — | — | — | 66.8 |