Captioning on RGBE-Chat
4.03CIRE-VLM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| RE-VLMInput=RGB+Event, Params=4B, Fine-tuned=true2026.05 | 4.03 | 3.5 | 4.34 | |
| Qwen2.5-VL*Input=RGB-only, Params=3B, Fine-tuned=true2026.05 | 3.91 | 3.41 | 4.27 | |
| DeepSeek2-VLInput=RGB-only, Params=7B, Fine-tuned=false2026.05 | 3.63 | 2.89 | 4.11 | |
| Intern2VLInput=RGB-only, Params=4B, Fine-tuned=false2026.05 | 3.47 | 2.84 | 3.91 | |
| Qwen2.5-VLInput=RGB-only, Params=3B, Fine-tuned=false2026.05 | 3.34 | 2.7 | 3.64 | |
| LLaVA-1.5Input=RGB-only, Params=7B, Fine-tuned=false2026.05 | 3.12 | 2.31 | 3.6 | |
| EventGPTInput=Event-only, Params=7B, Fine-tuned=false2026.05 | 2.82 | 2.34 | 3.08 | |
| Qwen2.5-VL*Input=Event-only, Params=3B, Fine-tuned=true2026.05 | 2.79 | 2.57 | 3.16 |