Object-Counting on Extreme Illumination Triplet Instruct-Following Dataset
74.66AccuracyOurs-7B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Ours-7BMLLM Type=Event-Enhanced Model, Backbone=Qwen-7B2026.03 | 74.66 | 0.4557 | |
| Baseline-7BMLLM Type=Event-Enhanced Model, Backbone=Qwen-7B2026.03 | 72.73 | 0.5303 | |
| InternVLMLLM Type=General Model2026.03 | 72.56 | 0.8641 | |
| Ours-3BMLLM Type=Event-Enhanced Model, Backbone=Qwen-3B2026.03 | 71.08 | 0.6127 | |
| Baseline-3BMLLM Type=Event-Enhanced Model, Backbone=Qwen-3B2026.03 | 69.66 | 0.6655 | |
| LLaVA-7BMLLM Type=General Model2026.03 | 67.84 | 0.9957 | |
| EventGPTMLLM Type=Event-specialized Model2026.03 | 67.41 | 1.0015 | |
| Q-InstructMLLM Type=Illumination-aware Model2026.03 | 66.87 | 0.9803 | |
| Deepseek-VLMLLM Type=General Model2026.03 | 54.87 | 1.0544 |