Visual World Modelling on Kubric
7.45GPT-4o ScoreSWIRL (ITER. + SHARE)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SWIRL (ITER. + SHARE)Method Strategy (SFT/Bootstrap/SWIRL)=SWIRL (ITER. + SHARE)2026.02 | 7.45 | 0.62 | 0.91 | |
| CHAMELEON-SFT2026.02 | 7.3 | — | — | |
| CHAMELEON-BOOTSTRAP2026.02 | 7.3 | — | — | |
| LIQUID-SFTMethod Strategy (SFT/Bootstrap/SWIRL)=SFT2026.02 | 7 | 0.6 | 0.9 | |
| SWIRL (ITERATIVE)Method Strategy (SFT/Bootstrap/SWIRL)=SWIRL (ITERATIVE)2026.02 | 6.96 | 0.54 | 0.9 | |
| UNIWORLD-V12026.02 | 6.76 | 0.42 | 0.84 | |
| SWIRL (IDM → FWM)Method Strategy (SFT/Bootstrap/SWIRL)=SWIRL (IDM → FWM)2026.02 | 6.59 | 0.5 | 0.9 | |
| LIQUID-SFTTest-time Verification (N)=82026.02 | 6.28 | 0.46 | 0.9 | |
| LIQUID-SFTTest-time Verification (N)=22026.02 | 6.18 | 0.44 | 0.9 | |
| LIQUID-SFTTest-time Verification (N)=42026.02 | 6.04 | 0.38 | 0.9 | |
| BAGEL-14B2026.02 | 5.82 | 0.18 | 0.94 | |
| OMNIGEN2026.02 | 5.69 | 0.16 | 0.91 | |
| LIQUID-BOOTSTRAPMethod Strategy (SFT/Bootstrap/SWIRL)=Bootstrap2026.02 | 5.57 | 0.3 | 0.9 | |
| OMNIGEN22026.02 | 5.54 | 0.14 | 0.89 | |
| GOT2026.02 | 3.92 | — | — | |
| SMARTEDIT2026.02 | 3.7 | — | — | |
| BLIP3O-NEXT2026.02 | 3.4 | 0.54 | 0.74 | |
| InstructPix2Pix2026.02 | 1.88 | — | — |