Video Question Answering on Video-MME (Overall, Short, Medium, Long Accuracy)
67.2Overall AccuracyMemoryCard (MiniCPM-V-4.5)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MemoryCard (MiniCPM-V-4.5)LLM Size=8B, #Frames=4+ 8+ 322026.06 | 67.2 | 76 | 67.6 | 58 | |
| MemoryCard (Qwen3-VL)LLM Size=8B, #Frames=4+ 8+ 322026.06 | 64.7 | 72.2 | 64.8 | 54.7 | |
| LongVULLM Size=7B, #Frames=*2026.06 | 60.9 | 64.7 | 58.2 | 59.5 | |
| MemoryCard (Qwen2-VL)LLM Size=7B, #Frames=4+ 8+ 322026.06 | 60.5 | 69.7 | 61 | 50.9 | |
| MiniCPM-V-4.5LLM Size=8B, #Frames=82026.06 | 59.9 | 69.6 | 58.2 | 52 | |
| MiniCPM-V-4.5-VideoLLM Size=8B, #Frames=82026.06 | 59.7 | 68.4 | 58.9 | 51.8 | |
| LLaVA-OneVisionLLM Size=7B, #Frames=*2026.06 | 58.2 | — | — | — | |
| Frame-VoyagerLLM Size=8B, #Frames=82026.06 | 57.5 | 67.3 | 56.3 | 48.9 | |
| Qwen3-VLLLM Size=8B, #Frames=82026.06 | 57.4 | 68.8 | 53.1 | 50.2 | |
| Qwen3-VL-VideoLLM Size=8B, #Frames=82026.06 | 56.6 | 67.7 | 51.7 | 50.3 | |
| Video-XLLLM Size=7B, #Frames=128/2562026.06 | 55.5 | 64 | 53.2 | 49.2 | |
| Qwen2-VLLLM Size=7B, #Frames=82026.06 | 53.7 | 65 | 50.7 | 45.3 | |
| Qwen2-VL-VideoLLM Size=7B, #Frames=82026.06 | 53 | 64.1 | 49.3 | 45.6 | |
| Video-CCAMLLM Size=9B, #Frames=962026.06 | 50.3 | 61.9 | 49.2 | 39.6 | |
| MovieChat+LLM Size=7B, #Frames=*2026.06 | 44.5 | 49.3 | 44.5 | 39.7 | |
| Qwen-VLLLM Size=7B, #Frames=82026.06 | 41.1 | 46.9 | 38.7 | 37.8 | |
| Video-LLaVALLM Size=7B, #Frames=82026.06 | 39.9 | 45.3 | 38 | 36.2 | |
| ShareGPT4VideoLLM Size=8B, #Frames=162026.06 | 39.9 | 48.3 | 36.3 | 35 | |
| VideoChat2LLM Size=7B, #Frames=82026.06 | 39.5 | 48.3 | 37 | 33.2 | |
| MovieChatLLM Size=7B, #Frames=*2026.06 | 38.2 | — | — | 33.4 |