Embodied Question Answering on BridgeEQA (test)
88.9Image Citation RelevanceEMVR VLM w/ Images + SG
Evaluation Results
| Method | Links | |
|---|---|---|
| EMVR VLM w/ Images + SGVLM=Grok 4 Fast2025.11 | 88.9 | |
| EMVR VLM w/ SG OnlyVLM=Grok 4 Fast2025.11 | 87.6 | |
| EMVR VLM w/ Images + SGVLM=Gemini 2.5 Flash Lite2025.11 | 84.9 | |
| Socratic LLM w/ SGVLM=Grok 4 Fast2025.11 | 83.8 | |
| EMVR VLM w/ SG OnlyVLM=Gemini 2.5 Flash2025.11 | 83.6 | |
| Multi-Frame VLM w/ SGVLM=Grok 4 Fast2025.11 | 83.3 | |
| EMVR VLM w/ Images + SGVLM=Gemini 2.5 Flash2025.11 | 80.3 | |
| Multi-Frame VLM w/ SGVLM=Gemini 2.5 Flash Lite2025.11 | 80.2 | |
| EMVR VLM w/ SG OnlyVLM=Gemini 2.5 Flash Lite2025.11 | 79.8 | |
| Multi-Frame VLM w/ SGVLM=Gemini 2.5 Flash2025.11 | 77.8 | |
| Socratic LLM w/ SGVLM=Gemini 2.5 Flash Lite2025.11 | 77.5 | |
| Socratic LLM w/ SGVLM=Gemini 2.5 Flash2025.11 | 76.7 | |
| Multi-Frame VLMVLM=Gemini 2.5 Flash Lite2025.11 | 71.7 | |
| Multi-Frame VLMVLM=Gemini 2.5 Flash2025.11 | 69.4 | |
| Multi-Frame VLMVLM=Grok 4 Fast2025.11 | 68.7 |