Multimodal Named Entity Recognition on MNER-MI
85.1F1 ScoreQwen2.5VL-7B + MRSI + CVO
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen2.5VL-7B + MRSI + CVOBackbone=Qwen2.5VL-7B, Training Configuration=MRSI + Constraint-guided Verifiable Optimization (CVO)2026.02 | 85.1 | 84.1 | 86.2 | — | |
| MimoVL-7B + MRSI + CVOBackbone=MimoVL-7B, Training Configuration=MRSI + Constraint-guided Verifiable Optimization (CVO)2026.02 | 84.8 | 84.7 | 85 | — | |
| Qwen2.5VL-7B + SFTBackbone=Qwen2.5VL-7B, Training Configuration=Supervised Fine-Tuning (SFT)2026.02 | 83.8 | 84 | 83.6 | — | |
| MimoVL-7B + MRSIBackbone=MimoVL-7B, Training Configuration=Multi-style Reasoning Schema Injection (MRSI)2026.02 | 83.3 | 83.2 | 83.4 | — | |
| Qwen2.5VL-7B + MRSIBackbone=Qwen2.5VL-7B, Training Configuration=Multi-style Reasoning Schema Injection (MRSI)2026.02 | 82 | 82.1 | 81.8 | — | |
| MimoVL-7B + SFTBackbone=MimoVL-7B, Training Configuration=Supervised Fine-Tuning (SFT)2026.02 | 81.23 | 81.8 | 80.8 | — | |
| SAVER2026.05 | 77.3 | — | — | 0.4 | |
| GLRA-adapt2026.05 | 76.9 | — | — | — |