Video Question Answering on HourVideo v1.0 (test)
33.4Overall AccuracyFocusGraph-m
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| FocusGraph-mQA LLM=Qwen-2.5-VL-7B, Frame Number=82026.03 | 33.4 | 21.7 | 32.6 | 33.6 | 44.3 | |
| GenSQA LLM=Qwen-2.5-VL-7B, Frame Number=82026.03 | 32.9 | 27.5 | 34.5 | 31.8 | 37.8 | |
| FocusGraph-sQA LLM=Qwen-2.5-VL-7B, Frame Number=82026.03 | 32.3 | 24.2 | 31.1 | 32.1 | 43.1 | |
| ViaRLQA LLM=Qwen-2.5-VL-7B, Frame Number=82026.03 | 32.23 | 17.5 | 33.4 | 31 | 43.9 | |
| Uniform SamplingQA LLM=Qwen-2.5-VL-7B, Frame Number=162026.03 | 31.22 | 15 | 36.8 | 27.5 | 43.9 | |
| Uniform SamplingQA LLM=Qwen-2.5-VL-7B, Frame Number=82026.03 | 29.86 | 10 | 33.7 | 27.9 | 37.8 | |
| MaxInfoQA LLM=Qwen-2.5-VL-7B, Frame Number=162026.03 | 29.61 | 17.5 | 33.9 | 26.6 | 40.2 | |
| ReMEmbRQA LLM=Llama3.1-8B, Frame Number=8¹2026.03 | 24.28 | 17.5 | 24.54 | 27.47 | 0 |