End-to-end Question Answering on MultimodalQA (test)
44.57EMLILaC (w/ MM-Embed)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LILaC (w/ MM-Embed)MLLM=Qwen2.5-VL 7B, In-domain fine-tuned=false2026.02 | 44.57 | 51.97 | |
| LILaC (w/ UniME)MLLM=Qwen2.5-VL 7B, In-domain fine-tuned=false2026.02 | 43.42 | 49.72 | |
| LILaC (w/ mmE5)MLLM=Qwen2.5-VL 7B, In-domain fine-tuned=false2026.02 | 40.72 | 47.46 | |
| NV-Embed-v2MLLM=Qwen2.5-VL 7B, In-domain fine-tuned=false2026.02 | 37.23 | 43.85 | |
| VisRAG-RetMLLM=MiniCPM V2.6, In-domain fine-tuned=true2026.02 | 28.18 | 34.01 | |
| ColPaliMLLM=Qwen2.5-VL 7B, In-domain fine-tuned=true2026.02 | 23.59 | 27.37 | |
| VisRAG-RetMLLM=Qwen2.5-VL 7B, In-domain fine-tuned=true2026.02 | 22.24 | 25.55 |