Text-intensive Visual Question Answering on VQAInfo
25.9ScoreLBR (ours)
Evaluation Results
| Method | Links | |
|---|---|---|
| LBR (ours)Backbone=LLaVA-NEXT, Parameter Scale=3B, LBR Mitigation=true2026.05 | 25.9 | |
| LLaVA-NEXT-3BBackbone=LLaVA-NEXT, Parameter Scale=3B, LBR Mitigation=false2026.05 | 25 | |
| LBR (ours)Backbone=LLaVA-1.5, Parameter Scale=13B, LBR Mitigation=true2026.05 | 23.9 | |
| LLaVA-1.5-13BBackbone=LLaVA-1.5, Parameter Scale=13B, LBR Mitigation=false2026.05 | 23.7 | |
| LBR (ours)Backbone=LLaVA-1.5, Parameter Scale=7B, LBR Mitigation=true2026.05 | 21.7 | |
| LLaVA-1.5-7BBackbone=LLaVA-1.5, Parameter Scale=7B, LBR Mitigation=false2026.05 | 21.5 |