Scientific Table & Figure Analysis on ANABENCH (test)
61.06Cosine SimilarityANAGENT
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| ANAGENTEvaluation Protocol=One-Shot, Backbone=Gemini-2.5-Flash, Size=-2026.02 | 61.06 | 61.34 | 20.52 | 17.4 | 4.06 | 11.47 | 47.64 | 10.98 | 29.31 | |
| ANAGENTEvaluation Protocol=One-Shot, Backbone=GPT-4.1-mini, Size=-2026.02 | 60.87 | 63.28 | 24.26 | 20.65 | 5.73 | 12.55 | 49.47 | 12.98 | 31.22 | |
| ANAGENTEvaluation Protocol=One-Shot, Backbone=Qwen3-VL, Size=8B2026.02 | 60.55 | 62.27 | 24.65 | 20.06 | 5.92 | 12.95 | 49.15 | 12.98 | 31.07 | |
| ANAGENTEvaluation Protocol=One-Shot, Backbone=InternVL-3.5, Size=8B2026.02 | 60.26 | 62.12 | 23.18 | 19.14 | 4.56 | 12.97 | 48.52 | 12.22 | 30.37 | |
| ANAGENTEvaluation Protocol=One-Shot, Backbone=Qwen2.5-VL, Size=7B2026.02 | 60.24 | 61 | 23.41 | 19.41 | 4.98 | 12.47 | 48.22 | 12.29 | 30.25 | |
| ANAGENTEvaluation Protocol=Zero-Shot, Backbone=GPT-4.1-mini, Size=-2026.02 | 59.94 | 61.63 | 22.75 | 18.19 | 4.81 | 12.26 | 48.11 | 11.75 | 29.93 | |
| ANAGENTEvaluation Protocol=Zero-Shot, Backbone=Qwen3-VL, Size=8B2026.02 | 59.76 | 61.53 | 23.07 | 17.75 | 4.98 | 12.2 | 48.12 | 11.64 | 29.88 | |
| ANAGENTEvaluation Protocol=One-Shot, Backbone=Qwen3-VL, Size=4B2026.02 | 59.64 | 60.61 | 22.42 | 18.05 | 4.03 | 11.51 | 47.55 | 11.2 | 29.38 | |
| ANAGENTEvaluation Protocol=Zero-Shot, Backbone=InternVL-3.5, Size=8B2026.02 | 59.46 | 61.25 | 22.59 | 17 | 3.88 | 11.68 | 47.77 | 10.85 | 29.31 | |
| ANAGENTEvaluation Protocol=Zero-Shot, Backbone=Qwen3-VL, Size=4B2026.02 | 59.41 | 60.21 | 21.23 | 16.27 | 3.9 | 11.33 | 46.95 | 10.5 | 28.73 | |
| ANAGENTEvaluation Protocol=One-Shot, Backbone=InternVL-3.5, Size=4B2026.02 | 59.11 | 60.52 | 22.6 | 18.04 | 3.82 | 11.5 | 47.41 | 11.12 | 29.27 | |
| ANAGENTEvaluation Protocol=Zero-Shot, Backbone=Qwen2.5-VL, Size=7B2026.02 | 58.91 | 60.41 | 21.59 | 17.47 | 4.11 | 11.85 | 46.97 | 11.14 | 29.06 | |
| ANAGENTEvaluation Protocol=One-Shot, Backbone=Qwen2.5-VL, Size=3B2026.02 | 58.89 | 60.7 | 22.19 | 18.41 | 3.99 | 11.54 | 47.26 | 11.31 | 29.29 | |
| ANAGENTEvaluation Protocol=Zero-Shot, Backbone=InternVL-3.5, Size=4B2026.02 | 58.26 | 59.86 | 21.21 | 16.1 | 3.29 | 11.11 | 46.44 | 10.17 | 28.31 | |
| ANAGENTEvaluation Protocol=Zero-Shot, Backbone=Qwen2.5-VL, Size=3B2026.02 | 57.5 | 60.01 | 21.03 | 17.34 | 3.87 | 11.53 | 46.18 | 10.91 | 28.55 | |
| GPT-4.1-miniEvaluation Protocol=Baseline, Size=-2026.02 | 56.34 | 59.74 | 19.47 | 16.74 | 3.39 | 11.49 | 45.18 | 10.54 | 27.86 | |
| Qwen3-VLEvaluation Protocol=Baseline, Size=8B2026.02 | 55.94 | 59.11 | 19.16 | 17.06 | 3.02 | 10.39 | 44.73 | 10.16 | 27.44 | |
| InternVL-3.5Evaluation Protocol=Baseline, Size=8B2026.02 | 55.73 | 59.1 | 19.3 | 16.8 | 2.86 | 10.28 | 44.71 | 9.98 | 27.34 | |
| Qwen2.5-VLEvaluation Protocol=Baseline, Size=7B2026.02 | 55.65 | 59.66 | 18.9 | 16.4 | 2.98 | 10.38 | 44.74 | 9.98 | 27.31 | |
| ANAGENTEvaluation Protocol=Zero-Shot, Backbone=Gemini-2.5-Flash, Size=-2026.02 | 55.6 | 59.37 | 19.4 | 16.04 | 3.15 | 11.1 | 44.79 | 10.09 | 27.44 | |
| Qwen3-VLEvaluation Protocol=Baseline, Size=4B2026.02 | 55.41 | 58.15 | 18.41 | 15.77 | 2.77 | 10.06 | 43.99 | 9.53 | 26.76 | |
| Qwen2.5-VLEvaluation Protocol=Baseline, Size=3B2026.02 | 54.74 | 58.49 | 17.82 | 15.89 | 2.56 | 10.02 | 43.68 | 9.49 | 26.59 | |
| InternVL-3.5Evaluation Protocol=Baseline, Size=4B2026.02 | 54.38 | 58.19 | 18.76 | 15.67 | 2.66 | 9.8 | 43.78 | 9.37 | 26.58 | |
| Gemini-2.5-FlashEvaluation Protocol=Baseline, Size=-2026.02 | 52.41 | 55.99 | 19.01 | 14.9 | 2.76 | 9.95 | 42.47 | 9.2 | 25.84 |