Conversational Response Generation on PRIST 1.0 (test)
0.1121BLEU-4OMG-LLaVA
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| OMG-LLaVAEvaluation Protocol=Fine-tuning, Model Type=Segmentation-specific MLLM2025.02 | 0.1121 | 0.123 | 0.353 | 0.3059 | 0.3918 | 0.8876 | |
| PixelLMEvaluation Protocol=Fine-tuning, Model Type=Segmentation-specific MLLM2025.02 | 0.0997 | 0.116 | 0.38 | 0.3063 | 0.3299 | 0.878 | |
| OMG-LLaVAEvaluation Protocol=Zero-shot, Model Type=Segmentation-specific MLLM2025.02 | 0.087 | 0.112 | 0.42 | 0.2347 | 0.279 | 0.873 | |
| MIRASEvaluation Protocol=Fine-tuning, Model Stage=Stage-22025.02 | 0.0851 | 0.157 | 0.492 | 0.3082 | 0.4006 | 0.8847 | |
| LISAEvaluation Protocol=Fine-tuning, Model Type=Segmentation-specific MLLM2025.02 | 0.0781 | 0.142 | 0.407 | 0.2784 | 0.3074 | 0.8675 | |
| GPT-4oEvaluation Protocol=Zero-shot, Model Type=General MLLM2025.02 | 0.043 | 0.091 | 0.429 | 0.2635 | 0.2855 | 0.8762 | |
| MIRASEvaluation Protocol=Zero-shot, Model Stage=Stage-12025.02 | 0.0417 | 0.069 | 0.284 | 0.2594 | 0.2818 | 0.8754 | |
| LLaVA-v1.5-7BEvaluation Protocol=Zero-shot, Model Type=General MLLM2025.02 | 0.0321 | 0.056 | 0.275 | 0.1602 | 0.1841 | 0.7881 | |
| Qwen2-VL-7BEvaluation Protocol=Zero-shot, Model Type=General MLLM2025.02 | 0.03 | 0.093 | 0.412 | 0.2628 | 0.2826 | 0.8673 | |
| LISAEvaluation Protocol=Zero-shot, Model Type=Segmentation-specific MLLM2025.02 | 0.0197 | 0.062 | 0.284 | 0.1821 | 0.2659 | 0.8567 | |
| PixelLMEvaluation Protocol=Zero-shot, Model Type=Segmentation-specific MLLM2025.02 | 0.0134 | 0.067 | 0.301 | 0.1493 | 0.2126 | 0.8513 | |
| InternVL2-8BEvaluation Protocol=Zero-shot, Model Type=General MLLM2025.02 | 0.012 | 0.025 | 0.41 | 0.1844 | 0.2398 | 0.8478 | |
| LLaVA-v1.6-7BEvaluation Protocol=Zero-shot, Model Type=General MLLM2025.02 | 0.0107 | 0.055 | 0.275 | 0.206 | 0.25 | 0.8426 |