Semantic Reasoning on LLATISA OOD
67Series Comparison AccuracyLLATISA
Evaluation Results
| Method | Links | |
|---|---|---|
| LLATISAModality=Vision(plot+num)2026.04 | 67 | |
| InternVL3-8BModality=Vision (plot+num)2026.04 | 58 | |
| InternVL3-8BModality=Vision + Text(w/o index)2026.04 | 56 | |
| InternVL3-8BModality=Vision + Text(w/ index)2026.04 | 55 | |
| Gemma3-12BModality=Vision + Text(w/ index)2026.04 | 53 | |
| Gemma3-12BModality=Vision(plot)2026.04 | 51 | |
| Gemma3-12BModality=Vision + Text(w/o index)2026.04 | 49 | |
| InternVL3-8BModality=Vision(plot)2026.04 | 46 | |
| Gemma3-12BModality=Vision (plot+num)2026.04 | 41 |