Metric: Avg. Accuracy (higher is better)
| # | Model↕ | Avg. Accuracy▼ | Extra Data | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | Human (Amateur) | 91.42 | No | Bongard-HOI: Benchmarking Few-Shot Visual Reason... | 2022-05-27 | Code |
| 2 | GPT-4o + CA | 77.3 | No | A Cognitive Paradigm Approach to Probe the Perce... | 2025-01-23 | - |
| 3 | SVM-Mimic + PMF (fine-tuned CLIP RN-50) | 76.41 | Yes | Support-Set Context Matters for Bongard Problems | 2023-09-07 | Code |
| 4 | Gemini 2.0 + CA | 74.5 | No | A Cognitive Paradigm Approach to Probe the Perce... | 2025-01-23 | - |
| 5 | SVM-Mimic (frozen CLIP RN-50) | 72.45 | Yes | Support-Set Context Matters for Bongard Problems | 2023-09-07 | Code |
| 6 | Meta-Baseline (ImagNet_R50) | 55.82 | No | Bongard-HOI: Benchmarking Few-Shot Visual Reason... | 2022-05-27 | Code |
| 7 | Meta-Baseline (MoCov2_R50) | 54.3 | No | Bongard-HOI: Benchmarking Few-Shot Visual Reason... | 2022-05-27 | Code |
| 8 | Meta-Baseline (Scratch_R50) | 54.23 | No | Bongard-HOI: Benchmarking Few-Shot Visual Reason... | 2022-05-27 | Code |
| 9 | ANIL (ImageNet_R50) | 49.74 | No | Bongard-HOI: Benchmarking Few-Shot Visual Reason... | 2022-05-27 | Code |