Regression on SDGBiasBench
8.93MAECADE (LLAVA-ONEVISION)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| CADE (LLAVA-ONEVISION)Params=7B, Setting=CADE2026.05 | 8.93 | -10.79 | 25 | 25 | |
| CADE (LLAVA-v1.5)Params=13B, Setting=CADE2026.05 | 9.16 | -10.55 | 43 | 16.22 | |
| CADE (INSTRUCTBLIP)Params=13B, Setting=CADE2026.05 | 9.71 | 10 | 37 | 12.12 | |
| CADE (INTERNVL3)Params=2B, Setting=CADE2026.05 | 9.81 | -5.03 | 37 | 27.59 | |
| CADE (LLAVA-v1.5)Params=7B, Setting=CADE2026.05 | 9.88 | -5.36 | 35 | 6.06 | |
| LLAVA-ONEVISIONParams=7B, Setting=Original2026.05 | 10.01 | — | 20 | — | |
| LLAVA-v1.5Params=13B, Setting=Original2026.05 | 10.24 | — | 37 | — | |
| INTERNVL3Params=2B, Setting=Original2026.05 | 10.33 | — | 29 | — | |
| LLAVA-v1.5Params=7B, Setting=Original2026.05 | 10.44 | — | 33 | — | |
| CADE (QWEN2.5-VL)Params=7B, Setting=CADE2026.05 | 10.45 | -16.27 | 31 | 40.91 | |
| INSTRUCTBLIPParams=13B, Setting=Original2026.05 | 10.8 | — | 33 | — | |
| GPT-4OSetting=API2026.05 | 11.25 | — | 25 | — | |
| CADE (INSTRUCTBLIP)Params=7B, Setting=CADE2026.05 | 11.45 | -49.31 | 20 | 81.82 | |
| CLAUDE 3.5 SONNETSetting=API2026.05 | 11.46 | — | 15 | — | |
| CADE (INTERNVL3)Params=14B, Setting=CADE2026.05 | 11.6 | -51.69 | 25 | 31.58 | |
| CADE (QWEN2.5-VL)Params=3B, Setting=CADE2026.05 | 11.93 | -1.32 | 28 | 3.7 | |
| QWEN2.5-VLParams=3B, Setting=Original2026.05 | 12.09 | — | 27 | — | |
| GEMINI 2.0 FLASHSetting=API2026.05 | 12.35 | — | 12 | — | |
| QWEN2.5-VLParams=7B, Setting=Original2026.05 | 12.48 | — | 22 | — | |
| CADE (INTERNVL3)Params=8B, Setting=CADE2026.05 | 14.42 | -27.46 | 20 | 42.86 | |
| INTERNVL3Params=8B, Setting=Original2026.05 | 19.88 | — | 14 | — | |
| INSTRUCTBLIPParams=7B, Setting=Original2026.05 | 22.59 | — | 11 | — | |
| INTERNVL3Params=14B, Setting=Original2026.05 | 24.01 | — | 19 | — |