Embodied Question Answering on Open-EQA (test)
52.8Object RecognitionFine-EQA w/ SFT+RL (VG-AVS)
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Fine-EQA w/ SFT+RL (VG-AVS)Method Component=SFT+RL, Framework=Visually-Grounded Active View Selection2025.12 | 52.8 | 65.2 | 61.4 | 58.2 | 36.6 | 46.2 | 53.4 | |
| Fine-EQA w/ SFT (VG-AVS)Method Component=SFT, Framework=Visually-Grounded Active View Selection2025.12 | 44.8 | 42.11 | 60 | 43.64 | 35.43 | 45.88 | 45.31 | |
| Fine-EQA w/ RL (VG-AVS)Method Component=RL, Framework=Visually-Grounded Active View Selection2025.12 | 41.6 | 46.32 | 55.56 | 43.03 | 35.43 | 48.24 | 45.03 | |
| Fine-EQAbaseline=Fine-EQA2025.12 | 40 | 45.26 | 51.11 | 52.73 | 41.71 | 48.24 | 46.51 |