Open-world image classification Very fine-grained set
24.9TISpeciaRL-7B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SpeciaRL-7BMethod Category=Reasoning LMMs - fine-tuned out-of-distribution2026.03 | 24.9 | 63.8 | 50.5 | 75.4 | |
| Qwen2.5VL-7B (rft)Method Category=Reasoning LMMs - fine-tuned out-of-distribution2026.03 | 21.9 | 68.2 | 49.5 | 74 | |
| Qwen2.5VL-7B (sft)Method Category=Reasoning LMMs - fine-tuned out-of-distribution2026.03 | 17.1 | 71.1 | 47.3 | 71.1 | |
| QWEN2.5VL 7BMethod Category=Reasoning LMMs2026.03 | 16.4 | 70.4 | 45.8 | 68.4 | |
| QWEN2VL 2BMethod Category=Non-reasoning LMMs2026.03 | 12.9 | 60.7 | 45.1 | 62.3 | |
| QWEN2.5VL 3BMethod Category=Reasoning LMMs2026.03 | 9.4 | 58.9 | 39.9 | 58.5 | |
| CLIP retrievalMethod Category=Retrieval-based baselines2026.03 | 7 | 18.1 | 39.7 | 56.1 | |
| INTERNVL2 8BMethod Category=Non-reasoning LMMs2026.03 | 2.3 | 32.5 | 36 | 49.4 | |
| INTERNVL2 4BMethod Category=Non-reasoning LMMs2026.03 | 1.7 | 36.8 | 33.8 | 44.2 | |
| INTERNVL2.5 8BMethod Category=Reasoning LMMs2026.03 | 1.6 | 59.9 | 32.1 | 40.4 | |
| LLAVA-NEXT Mist 7BMethod Category=Non-reasoning LMMs2026.03 | 1.4 | 47.2 | 34.2 | 46.9 | |
| LLAVA-NEXT Vic 7BMethod Category=Non-reasoning LMMs2026.03 | 1.3 | 42.2 | 34.5 | 46.1 | |
| QWEN2VL 7BMethod Category=Non-reasoning LMMs2026.03 | 0.8 | 63 | 34.5 | 43.4 | |
| INTERNVL2.5 2BMethod Category=Reasoning LMMs2026.03 | 0.8 | 52.4 | 31.6 | 41.5 | |
| CASEDMethod Category=Retrieval-based baselines2026.03 | 0.7 | 47.1 | 38.5 | 38.5 | |
| INTERNVL2 2BMethod Category=Non-reasoning LMMs2026.03 | 0.7 | 32.9 | 33.1 | 43.9 | |
| LLAVA-OV Qwen2 0.5BMethod Category=Non-reasoning LMMs2026.03 | 0.6 | 65.6 | 30.5 | 37.1 | |
| INTERNVL2.5 4BMethod Category=Reasoning LMMs2026.03 | 0.5 | 55.6 | 31.4 | 39.7 | |
| PHI-3-VISIONMethod Category=Non-reasoning LMMs2026.03 | 0.2 | 45 | 28.9 | 36 | |
| IDEFICS2 8BMethod Category=Non-reasoning LMMs2026.03 | 0 | 67 | 29.6 | 33.6 | |
| INSTRUCTBLIP Vic 7BMethod Category=Non-reasoning LMMs2026.03 | 0 | 61 | 30 | 34.3 | |
| LLAVA-1.5 7BMethod Category=Non-reasoning LMMs2026.03 | 0 | 41 | 28.6 | 37.6 | |
| LLAVA-OV Qwen2 7BMethod Category=Non-reasoning LMMs2026.03 | 0 | 76.7 | 31.9 | 32.4 |