Bongard Problem Solving on Bongard OpenWorld
95.6Generative ScoreLoRA (Lcombined)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| LoRA (Lcombined)Model=Phi2025.07 | 95.6 | 84.4 | 93.8 | |
| LoRA (Lcombined)Model=Gemma3 4B¹2025.07 | 95.6 | 89.8 | 96.6 | |
| LoRA (Lcombined)Model=Pixtral2025.07 | 95 | 74.4 | 96.2 | |
| LoRA (LNT)Model=Pixtral2025.07 | 93.4 | 76.6 | 87.2 | |
| LoRA (LNT)Model=Gemma3 4B¹2025.07 | 92.4 | 89.8 | 50 | |
| LoRA (LNT)Model=Phi2025.07 | 92.2 | 84.4 | 74.8 | |
| Human2025.07 | 91 | — | — | |
| Direct baselineModel=Gemma3 4B¹2025.07 | 76 | 89.8 | 50 | |
| Direct baselineModel=Pixtral2025.07 | 72.4 | 76 | 87.2 | |
| Direct baselineModel=Phi2025.07 | 59 | 84 | 76.4 |