Hallucination Assessment on MMHalu
4.27MMHalu ScoreLEAD
Evaluation results
| Method | Links | |
|---|---|---|
| LEADBase Model=VL-Rethinker-7B, Decoding Strategy=LEAD2026.03 | 4.27 | |
| LEADBase Model=VL-Cogito-7B, Decoding Strategy=LEAD2026.03 | 4.13 | |
| V-STARData Size=40k2026.04 | 4.12 | |
| VL-Rethinker-7BDecoding Strategy=Standard2026.03 | 4.06 | |
| VL-RethinkerData Size=39k2026.04 | 4.06 | |
| VL-Cogito-7BDecoding Strategy=Standard2026.03 | 3.95 | |
| VL-CogitoData Size=80k2026.04 | 3.95 | |
| LEADBase Model=Vision-R1-7B, Decoding Strategy=LEAD2026.03 | 3.89 | |
| LEADBase Model=R1-Onevision-7B, Decoding Strategy=LEAD2026.03 | 3.8 | |
| LEADBase Model=OpenVLThinker-7B, Decoding Strategy=LEAD2026.03 | 3.76 | |
| SIDBase Model=R1-Onevision-7B, Decoding Strategy=SID2026.03 | 3.7 | |
| MemVRBase Model=R1-Onevision-7B, Decoding Strategy=MemVR2026.03 | 3.69 | |
| Vision-R1-7BDecoding Strategy=Standard2026.03 | 3.64 | |
| OpenVLThinker-7BDecoding Strategy=Standard2026.03 | 3.59 | |
| Vision-R1Data Size=210k2026.04 | 3.57 | |
| VCDBase Model=R1-Onevision-7B, Decoding Strategy=VCD2026.03 | 3.55 | |
| R1-Onevision-7BDecoding Strategy=Standard2026.03 | 3.52 | |
| Qwen2.5VL2026.04 | 3.5 | |
| ThinkLite-VLData Size=11k2026.04 | 3.3 | |
| R1-OnevisionData Size=155k2026.04 | 3.26 | |
| OpenVLThinkerData Size=59.2k2026.04 | 3 |