Science Question Answering on ScienceQA (AUC)
92.36AUCReLope
Evaluation Results
| Method | Links | |
|---|---|---|
| ReLopeBackbone=Phi-4-Multimodal-Instruct2026.03 | 92.36 | |
| ReLopeBackbone=Qwen2.5-VL-7B-Instruct2026.03 | 91.79 | |
| ReLopeBackbone=Gemma3-12B2026.03 | 88.73 | |
| Attention ProbeBackbone=Phi-4-Multimodal-Instruct2026.03 | 87.24 | |
| Attention ProbeBackbone=Qwen2.5-VL-7B-Instruct2026.03 | 86.43 | |
| Attention ProbeBackbone=Gemma3-12B2026.03 | 84.93 | |
| RouteLLMBackbone=Phi-4-Multimodal-Instruct2026.03 | 84.13 | |
| ProbeBackbone=Qwen2.5-VL-7B-Instruct2026.03 | 83.95 | |
| RouteLLMBackbone=Qwen2.5-VL-7B-Instruct2026.03 | 82.83 | |
| ProbeBackbone=Phi-4-Multimodal-Instruct2026.03 | 82.08 | |
| ProbeBackbone=Gemma3-12B2026.03 | 80.47 | |
| RouteLLMBackbone=Gemma3-12B2026.03 | 78.92 | |
| Post-Hoc EmbedBackbone=Phi-4-Multimodal-Instruct2026.03 | 77.84 | |
| Cascade RoutingBackbone=Qwen2.5-VL-7B-Instruct2026.03 | 76.31 | |
| Post-Hoc EmbedBackbone=Gemma3-12B2026.03 | 76.19 | |
| Cascade RoutingBackbone=Gemma3-12B2026.03 | 75.58 | |
| Cascade RoutingBackbone=Phi-4-Multimodal-Instruct2026.03 | 75.56 | |
| Post-Hoc EmbedBackbone=Qwen2.5-VL-7B-Instruct2026.03 | 75.45 | |
| MoTBackbone=Phi-4-Multimodal-Instruct2026.03 | 74.03 | |
| MoTBackbone=Gemma3-12B2026.03 | 72.84 | |
| MoTBackbone=Qwen2.5-VL-7B-Instruct2026.03 | 72.08 |