Object Hallucination Evaluation on Object HalBench (test)
43.7CHAIRS ScoreSID
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SIDBackbone=R1-Onevision-7B2026.05 | 43.7 | 9.2 | |
| VCDBackbone=R1-Onevision-7B2026.05 | 42.3 | 9.4 | |
| OPA-DPOBackbone=R1-Onevision-7B2026.05 | 40 | 8.6 | |
| RLAIF-VBackbone=R1-Onevision-7B2026.05 | 38 | 7.9 | |
| VanillaBackbone=R1-Onevision-7B2026.05 | 36.7 | 7.6 | |
| SIDBackbone=OpenVLThinker-7B2026.05 | 35.7 | 8.9 | |
| SIDBackbone=MM-Eureka-7B2026.05 | 35.3 | 8.3 | |
| VCDBackbone=ThinkLite-VL-7B2026.05 | 35 | 8.2 | |
| SIDBackbone=ThinkLite-VL-7B2026.05 | 34.7 | 8.4 | |
| VanillaBackbone=MM-Eureka-7B2026.05 | 34 | 7.5 | |
| RLAIF-VBackbone=MM-Eureka-7B2026.05 | 33.3 | 7.3 | |
| RLAIF-VBackbone=ThinkLite-VL-7B2026.05 | 33 | 7.8 | |
| VanillaBackbone=OpenVLThinker-7B2026.05 | 31.7 | 7.3 | |
| VanillaBackbone=ThinkLite-VL-7B2026.05 | 30.7 | 7.5 | |
| OPA-DPOBackbone=ThinkLite-VL-7B2026.05 | 30.3 | 7.6 | |
| OPA-DPOBackbone=MM-Eureka-7B2026.05 | 29 | 5.4 | |
| VCDBackbone=MM-Eureka-7B2026.05 | 28.7 | 8.2 | |
| RLAIF-VBackbone=OpenVLThinker-7B2026.05 | 27.3 | 6.8 | |
| VCDBackbone=OpenVLThinker-7B2026.05 | 27 | 8.1 | |
| OPA-DPOBackbone=OpenVLThinker-7B2026.05 | 27 | 7.1 | |
| RC-DPOBackbone=OpenVLThinker-7B2026.05 | 24.3 | 6.4 | |
| RC-DPOBackbone=R1-Onevision-7B2026.05 | 21.7 | 6.2 | |
| RC-DPOBackbone=MM-Eureka-7B2026.05 | 21.3 | 5.4 | |
| RC-DPOBackbone=ThinkLite-VL-7B2026.05 | 18.7 | 4.9 |