Location Prediction on TARA (test)
38.4EM@1Qwen2.5VL-7B
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Qwen2.5VL-7BEvaluation Setup=Open-ended, Evidence Source=NewsRECON retrieved articles2026.01 | 38.4 | — | 61.6 | 40.3 | 58.7 | 60.4 | |
| InternVL3-8BEvaluation Setup=Open-ended, Evidence Source=NewsRECON retrieved articles2026.01 | 35.2 | — | 58.5 | 36.9 | 56 | 58.2 | |
| Qwen2.5VL-7BEvaluation Setup=Open-ended, Prompting Strategy=celebrity2026.01 | 32.8 | — | 57 | 34.5 | 51.6 | 55.7 | |
| Qwen2.5VL-7BEvaluation Setup=Open-ended, Prompting Strategy=default2026.01 | 32.5 | — | 56.7 | 34.2 | 50.5 | 54.9 | |
| Qwen2.5VL-7BEvaluation Setup=Open-ended, Prompting Strategy=GETReason2026.01 | 32.2 | — | 54 | 34 | 49.5 | 54 | |
| NewsRECONEvaluation Setup=Simplified2026.01 | 30.9 | 52.9 | 53.3 | 32.2 | 48.7 | 53.5 | |
| NewsRECONEvaluation Setup=Open-ended2026.01 | 29.1 | 41.7 | 55.5 | 30.8 | 49.5 | 54.3 | |
| InternVL3-8BEvaluation Setup=Open-ended, Prompting Strategy=celebrity2026.01 | 27.9 | — | 50.4 | 28.9 | 44.2 | 50.9 | |
| InternVL3-8BEvaluation Setup=Open-ended, Prompting Strategy=default2026.01 | 25.4 | — | 48.3 | 26.3 | 41.4 | 48.9 | |
| CLIP+Evaluation Setup=Simplified2026.01 | 19.4 | 35 | 38.7 | 19.7 | 32.3 | 43.8 | |
| InternVL3-8BEvaluation Setup=Open-ended, Prompting Strategy=GETReason2026.01 | 17.9 | — | 35.1 | 18.9 | 32.1 | 33 |