Knowledge-Based Visual Question Answering on E-VQA
12.5Final Fidelity RateLLaVA-1.5
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LLaVA-1.5LLM/MLLM Backbone=LLaMA-3.1-8B2026.05 | 12.5 | 16 | 16.9 | |
| LLaVA-1.5LLM/MLLM Backbone=Vicuna-7B2026.05 | 12.5 | 16.3 | 16.9 | |
| DPRV +TLLM/MLLM Backbone=Multi-passage BERT2026.05 | 11.2 | 29.1 | — | |
| Qwen2-VL-7BLLM/MLLM Backbone=-2026.05 | 11.1 | 16.2 | 17 | |
| Qwen2.5-VL-3BLLM/MLLM Backbone=-2026.05 | 10 | 17.9 | 19.6 | |
| RORA-VLMLLM/MLLM Backbone=Vicuna-7B2026.05 | 9.2 | — | 20.3 | |
| EchoSightLLM/MLLM Backbone=Mistral-7B/LLaMA-3-8B2026.05 | 9.1 | 19.4 | — | |
| EchoSightLLM/MLLM Backbone=LLaMA-3.1-8B2026.05 | 9.1 | 26.4 | 24.9 | |
| Qwen2.5-VL-7BLLM/MLLM Backbone=-2026.05 | 9 | 21.7 | 20.3 | |
| LLaVA-NeXT-7BLLM/MLLM Backbone=-2026.05 | 9 | 22.1 | 20.8 | |
| LLaVA-NeXT-8BLLM/MLLM Backbone=-2026.05 | 9 | 22.5 | 21.4 | |
| ReflectiVALLM/MLLM Backbone=LLaMA-3.1-8B2026.05 | 9 | 35.5 | 35.5 | |
| Wiki-LLaVALLM/MLLM Backbone=LLaMA-3.1-8B2026.05 | 8.3 | 18.3 | 19.6 | |
| Wiki-LLaVALLM/MLLM Backbone=Vicuna-7B2026.05 | 8.2 | 17.7 | 20.3 | |
| KU-RAGLLM/MLLM Backbone=GPT-4o2026.05 | 7.1 | 38.3 | — | |
| MMKB-RAGLLM/MLLM Backbone=Qwen2-VL-7B2026.05 | 6.6 | 39.7 | 35.9 | |
| VLM-PRFLLM/MLLM Backbone=Qwen2.5-VL-7B2026.05 | 6.4 | 28.9 | 28.6 | |
| VLM-PRF w/ RLLLM/MLLM Backbone=InternVL3-8B2026.05 | 5.8 | 40.1 | 39.2 |