Visual Question Answering on GQA (Score, Performance retention rate)
61.1ScoreUpper Bound
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Upper BoundTokens=5762025.12 | 61.1 | 100 | |
| HTC-VLMTokens=1 (hybrid)2025.12 | 57.6 | 85 | |
| VoCo-LLaMATokens=12025.12 | 57.4 | 84.2 | |
| Avg. PoolTokens=12025.12 | 52.9 | 65 | |
| Q-FormerTokens=12025.12 | 51.1 | 57.3 | |
| Lower BoundTokens=12025.12 | 37.7 | 0 |