Multi-page Question Answering on LongBench multi-page
48.9HQA F1AGAR
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| AGARBackbone=InternVL3.5-8B2026.06 | 48.9 | 43.8 | 25.6 | 38 | — | |
| Plain VQABackbone=InternVL3.5-8B2026.06 | 48.6 | 43.1 | 24 | 37.2 | — | |
| UniformBackbone=InternVL3.5-8B2026.06 | 48.6 | 43.1 | 24 | 37.2 | — | |
| VEA-BrBackbone=InternVL3.5-8B2026.06 | 48.3 | 31.3 | 7.1 | 21 | — | |
| RandomBackbone=InternVL3.5-8B2026.06 | 47.5 | 42.9 | 23.8 | 37.3 | — | |
| AGARBackbone=GLM-4.1V-9B-Thinking2026.06 | 46.9 | 50.7 | 26.8 | 25.9 | — | |
| VEA-BrBackbone=GLM-4.1V-9B-Thinking2026.06 | 46.7 | 50.2 | 24.7 | 23.6 | — | |
| VEA-ConBackbone=GLM-4.1V-9B-Thinking2026.06 | 45.9 | 49.9 | 23.9 | 24.9 | — | |
| CGRBackbone=GLM-4.1V-9B-Thinking2026.06 | 45.8 | 24.7 | 21.9 | 19.9 | — | |
| AGLABackbone=GLM-4.1V-9B-Thinking2026.06 | 45.7 | 38.9 | 26.5 | 18.6 | — | |
| VARBackbone=GLM-4.1V-9B-Thinking2026.06 | 45.6 | 50.3 | 24.5 | 22.7 | — | |
| Plain VQABackbone=GLM-4.1V-9B-Thinking2026.06 | 44.5 | 50.1 | 24.8 | 24.8 | — | |
| UniformBackbone=GLM-4.1V-9B-Thinking2026.06 | 44.5 | 50.2 | 24.8 | 24.2 | — | |
| AGARAvg. tok.=5,3842026.06 | 44.3 | 40.8 | 22.1 | 34.2 | — | |
| RandomBackbone=GLM-4.1V-9B-Thinking2026.06 | 43.5 | 46.5 | 25.5 | 23.9 | — | |
| VEA-ConAvg. tok.=4,9162026.06 | 42.1 | 36.1 | 18.2 | 29.9 | — | |
| VEA-BrAvg. tok.=4,9162026.06 | 41.8 | 35.2 | 21.3 | 30.3 | — | |
| AGLAAvg. tok.=4,9162026.06 | 41.2 | 39.1 | 21.7 | 31.5 | — | |
| RandomAvg. tok.=5,3512026.06 | 40.8 | 37.3 | 17.6 | 33.8 | — | |
| Plain VQAAvg. tok.=4,9162026.06 | 39.8 | 37.5 | 17.2 | 29.5 | — | |
| UniformAvg. tok.=5,3782026.06 | 39.8 | 37.5 | 16.9 | 33.3 | — | |
| VARAvg. tok.=4,9162026.06 | 37.7 | 39.2 | 18.5 | 30.9 | — | |
| VEA-ConBackbone=InternVL3.5-8B2026.06 | 37 | 32.9 | 10.1 | 22.4 | — | |
| CGRAvg. tok.=4,9162026.06 | 34.1 | 36.6 | 21.2 | 29.3 | — | |
| AGLABackbone=InternVL3.5-8B2026.06 | 31.7 | 29.5 | 17.8 | 17.9 | — | |
| VARBackbone=InternVL3.5-8B2026.06 | 30 | 30.2 | 13.4 | 12.4 | — | |
| CGRBackbone=InternVL3.5-8B2026.06 | 24.9 | 23.8 | 9.5 | 19.5 | — |