Comparative Relationship Prediction on PANDABENCH Easy
58AccuracyPANDA
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| PANDAModel Type=Open-source2026.04 | 58 | 61 | 54 | 56 | |
| DepictQAModel Type=Open-source, Trained on PANDASET=true2026.04 | 49 | 48 | 38 | 42 | |
| Attentive ProbeModel Type=Baseline2026.04 | 47 | 47 | 42 | 43 | |
| Linear ProbeModel Type=Baseline2026.04 | 37 | 35 | 22 | 15 | |
| GPT-5 NanoModel Type=Closed-source2026.04 | 34 | 26 | 28 | 26 | |
| GPT-5 MiniModel Type=Closed-source2026.04 | 31 | 32 | 31 | 26 | |
| GPT-4oModel Type=Closed-source2026.04 | 26 | 29 | 26 | 23 | |
| Gemini 2.5 ProModel Type=Closed-source2026.04 | 22 | 29 | 25 | 18 | |
| RandomModel Type=Baseline2026.04 | 20 | 20 | 20 | 19 |