Multimodal Reasoning on POPE
88.41AccuracyCFPOG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CFPOGBackbone=Qwen2.5-VL-7B, Hyperparameters=gamma=0.02 + No Ent2026.06 | 88.41 | — | |
| PAPOGBackbone=Qwen2.5-VL-7B2026.06 | 87.7 | — | |
| LaReBase VLM=Qwen3-VL-4B-Instruct2025.11 | 77.4 | 29 | |
| CoF*Base VLM=Qwen3-VL-4B-Instruct2025.11 | 71 | 164.1 | |
| LVR*Base VLM=Qwen3-VL-4B-Instruct2025.11 | 69.4 | 42.4 | |
| CCoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 69.1 | 112.8 | |
| LaReBase VLM=Qwen3-VL-2B-Instruct2025.11 | 68 | 18.9 | |
| ICoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 66.3 | 103.7 | |
| ICoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 65.9 | 104.7 | |
| CoF*Base VLM=Qwen3-VL-2B-Instruct2025.11 | 65.7 | 128.9 | |
| LVR*Base VLM=Qwen3-VL-2B-Instruct2025.11 | 64.5 | 29.5 | |
| MCOUTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 64.5 | 56 | |
| MCOUTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 64 | 37.2 | |
| CCoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 63.3 | 110.1 | |
| Standard CoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 58.2 | 104.2 | |
| MM-CoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 55.7 | 102.7 | |
| Standard CoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 55.1 | 99.5 | |
| MM-CoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 51.2 | 100.9 |