Multi-modal Reasoning on MMBench Overall & Relation Reasoning
84.7Overall AccuracyChainMPQ
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ChainMPQBackbone=Qwen2.5-VL-7B2025.10 | 84.7 | 1.5 | 81.8 | 3.6 | |
| ChainMPQBackbone=InternVL3-8B2025.10 | 84.2 | 0.6 | 83.9 | 1.4 | |
| InternVL3-8BBackbone=InternVL3-8B2025.10 | 83.6 | — | 82.5 | — | |
| Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2025.10 | 83.2 | — | 78.2 | — | |
| ChainMPQBackbone=LLaVA-v1.5-7B2025.10 | 67.8 | 1.3 | 61.3 | 2.5 | |
| LLaVA-v1.5-7BBackbone=LLaVA-v1.5-7B2025.10 | 66.5 | — | 58.8 | — | |
| ChainMPQBackbone=InstructBLIP-7B2025.10 | 65.5 | 1.6 | 55.2 | 2.7 | |
| InstructBLIP-7BBackbone=InstructBLIP-7B2025.10 | 63.9 | — | 52.5 | — |