Multi-page Memory-based Question Answering on LoCoMo multi-page memory
44.5SingleHop AccuracyAGAR
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| AGAR2026.06 | 44.5 | 30.7 | 23.2 | 24.3 | — | — | — | — | — | |
| VAR2026.06 | 43.2 | 29.4 | 22 | 21.5 | — | — | — | — | — | |
| Random2026.06 | 43 | 26.9 | 19 | 20.9 | — | — | — | — | — | |
| VEA-Con2026.06 | 43 | 30.4 | 22.2 | 18.2 | — | — | — | — | — | |
| Plain VQA2026.06 | 42 | 29.1 | 21.2 | 23 | — | — | — | — | — | |
| Uniform2026.06 | 42 | 29.1 | 21.2 | 23.1 | — | — | — | — | — | |
| VEA-Br2026.06 | 41.4 | 27.2 | 19.8 | 18.5 | — | — | — | — | — | |
| AGLA2026.06 | 38.5 | 28 | 21.9 | 18.6 | — | — | — | — | — | |
| CGR2026.06 | 36.5 | 27.3 | 15.7 | 14.6 | — | — | — | — | — | |
| AGARBackbone=GLM-4.1V-9B-Thinking2026.06 | 32.7 | 22 | 13.5 | 16.2 | — | — | — | — | — | |
| VARBackbone=GLM-4.1V-9B-Thinking2026.06 | 32.5 | 19.4 | 11.8 | 11.7 | — | — | — | — | — | |
| AGLABackbone=GLM-4.1V-9B-Thinking2026.06 | 32.2 | 17.4 | 9 | 7.7 | — | — | — | — | — | |
| AGARAvg. tok.=5,3842026.06 | 32 | 21.9 | 25.7 | 16.8 | — | — | — | — | — | |
| VEA-BrBackbone=GLM-4.1V-9B-Thinking2026.06 | 31 | 17.8 | 12.6 | 10.5 | — | — | — | — | — | |
| CGRAvg. tok.=4,9162026.06 | 30.1 | 21.8 | 21.2 | 15.8 | — | — | — | — | — | |
| VEA-ConBackbone=GLM-4.1V-9B-Thinking2026.06 | 29.8 | 18.3 | 12.1 | 13.3 | — | — | — | — | — | |
| VEA-BrAvg. tok.=4,9162026.06 | 29.4 | 20.3 | 23.5 | 15.4 | — | — | — | — | — | |
| Plain VQABackbone=GLM-4.1V-9B-Thinking2026.06 | 29.2 | 17.8 | 11.9 | 13.7 | — | — | — | — | — | |
| UniformBackbone=GLM-4.1V-9B-Thinking2026.06 | 29.2 | 17.8 | 11.9 | 13.7 | — | — | — | — | — | |
| RandomBackbone=GLM-4.1V-9B-Thinking2026.06 | 28.6 | 20.2 | 11.5 | 9.4 | — | — | — | — | — | |
| RandomAvg. tok.=5,3512026.06 | 28 | 19 | 22.6 | 14.8 | — | — | — | — | — | |
| VEA-ConAvg. tok.=4,9162026.06 | 27.8 | 19.2 | 23.1 | 13.5 | — | — | — | — | — | |
| CGRBackbone=GLM-4.1V-9B-Thinking2026.06 | 27.5 | 14.8 | 4.9 | 7.2 | — | — | — | — | — | |
| UniformAvg. tok.=5,3782026.06 | 26.5 | 18.5 | 21.8 | 12.2 | — | — | — | — | — | |
| Plain VQAAvg. tok.=4,9162026.06 | 26.1 | 20.5 | 22.7 | 12.1 | — | — | — | — | — | |
| AGLAAvg. tok.=4,9162026.06 | 26.1 | 19.1 | 24.8 | 15.2 | — | — | — | — | — | |
| VARAvg. tok.=4,9162026.06 | 24.3 | 17.1 | 20 | 16 | — | — | — | — | — | |
| AGARBackbone=InternVL3.5-8B2026.06 | — | — | — | — | 42.2 | 27.6 | 31.4 | 17.8 | — | |
| AGLABackbone=InternVL3.5-8B2026.06 | — | — | — | — | 15.3 | 14.6 | 16.6 | 14.9 | — | |
| CGRBackbone=InternVL3.5-8B2026.06 | — | — | — | — | 12.2 | 10.3 | 5.7 | 9.9 | — | |
| Plain VQABackbone=InternVL3.5-8B2026.06 | — | — | — | — | 35.9 | 26 | 28.1 | 16.4 | — | |
| RandomBackbone=InternVL3.5-8B2026.06 | — | — | — | — | 38.6 | 26.9 | 28.3 | 17.2 | — | |
| UniformBackbone=InternVL3.5-8B2026.06 | — | — | — | — | 35.8 | 26 | 28.1 | 16.4 | — | |
| VARBackbone=InternVL3.5-8B2026.06 | — | — | — | — | 11.6 | 13.5 | 15.6 | 11.8 | — | |
| VEA-BrBackbone=InternVL3.5-8B2026.06 | — | — | — | — | 15.3 | 13.1 | 31.4 | 8.1 | — | |
| VEA-ConBackbone=InternVL3.5-8B2026.06 | — | — | — | — | 17.2 | 15 | 31.1 | 8.1 | — |