Predictive Inference on L4 Out-of-Distribution (test)
83.3AccuracyLLaTiSA (L1→L2→L3→L4)
Evaluation Results
| Method | Links | |
|---|---|---|
| LLaTiSA (L1→L2→L3→L4)Modality=Vision (plot + num), Trained_on=L1 to L4 subset2026.04 | 83.3 | |
| Claude-3.5-SonnetModality=Text (w/o index)2026.04 | 82.2 | |
| GPT-4.1Modality=Text (w/o index)2026.04 | 79.1 | |
| GPT-4oModality=Vision (plot + num)2026.04 | 78.3 | |
| GPT-4oModality=Text (w/o index)2026.04 | 75.6 | |
| Qwen3-8BModality=Text (w/o index)2026.04 | 67.1 | |
| LLaTiSA (L1→L2→L3)Modality=Vision (plot + num), Trained_on=L1 to L3 subset2026.04 | 54.2 | |
| Qwen3-VL-8BModality=Vision (plot + num)2026.04 | 42.1 | |
| LLaMA3.1-8BModality=Text (w/o index)2026.04 | 41 |