Video Captioning on KPM-Cap 1.0 (test)
43.64BLEU-1KPM (w/o MoPE)
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| KPM (w/o MoPE)Utilized-LLM=Qwen2.5, Size=7B2026.02 | 43.64 | 15.14 | 34.11 | 96.57 | 3.39 | 67.83 | |
| KPMUtilized-LLM=Qwen2.5, Size=7B2026.02 | 43.25 | 14.77 | 33.4 | 96.51 | 3.46 | 71.8 | |
| GPT-4.1Utilized-LLM=Closed-source, Size=N/A2026.02 | 38.78 | 9.99 | 29.68 | 93.73 | 3.23 | 62.41 | |
| Tarsier2-RecapUtilized-LLM=Qwen2, Size=7B2026.02 | 33.58 | 5.61 | 23.99 | 93.62 | 3.01 | 43.51 | |
| Gemini-2.5ProUtilized-LLM=Closed-source, Size=N/A2026.02 | 32.22 | 4.52 | 25.52 | 92.8 | 3.11 | 58.79 | |
| LLaVA-OneVisionUtilized-LLM=Qwen2, Size=7B2026.02 | 29.77 | 3.24 | 21.07 | 93.74 | 2.53 | 31.55 | |
| Qwen2.5-VLUtilized-LLM=Qwen2.5, Size=7B2026.02 | 27.51 | 2.83 | 20.14 | 91.23 | 3.19 | 54.63 | |
| MiniCPM-V2.6Utilized-LLM=Qwen2, Size=8B2026.02 | 24.39 | 2.24 | 20.06 | 92.77 | 2.52 | 34.36 | |
| VideoLLaMA3Utilized-LLM=Qwen2.5, Size=7B2026.02 | 23.49 | 3.18 | 21.86 | 85.88 | 2.3 | 28.18 | |
| Qwen2-VLUtilized-LLM=Qwen2, Size=7B2026.02 | 21.6 | 1.74 | 19.19 | 89.65 | 2.6 | 25.64 | |
| InternVideo-2.5Utilized-LLM=InternLM2.5, Size=8B2026.02 | 20.46 | 3.05 | 19.45 | 86.56 | 2.35 | 30.7 |