Visual Question Answering on Infoseek M2KR setup (test)
42.8VQA AccuracyBEFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BEFTAnswer Generator=Qwen2-VL-7B-Instruct, Retriever=PreFLMR-L2026.04 | 42.8 | 36.6 | |
| EchoSight+Answer Generator=LLaMA-3-8B2026.04 | 41.8 | 74 | |
| MuKAAnswer Generator=VILA-13B2026.04 | 41 | 51.7 | |
| ReflectivaAnswer Generator=LLaMA-3.1-8B2026.04 | 40.1 | 77.6 | |
| DPOAnswer Generator=Qwen2-VL-7B-Instruct, Retriever=PreFLMR-L2026.04 | 35.3 | 36.6 | |
| RAVQA-v2Answer Generator=T5-Large2026.04 | 33.5 | 36.6 | |
| SFTAnswer Generator=Qwen2-VL-7B-Instruct, Retriever=PreFLMR-L2026.04 | 32.2 | 36.6 | |
| UniIRAnswer Generator=T5-Large2026.04 | 26.4 | 34.5 |