Content Description Quality on XM3600
65.28Preference RateLLaVA-PLLuM-12B-nc
Evaluation Results
| Method | Links | |
|---|---|---|
| LLaVA-PLLuM-12B-ncJudge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=PaliGemma2-10B2026.02 | 65.28 | |
| LLaVA-PLLuM-12B-nc-250715Judge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=PaliGemma2-10B2026.02 | 64.83 | |
| LLaVA-Bielik-11B-v2.6Judge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=PaliGemma2-10B2026.02 | 62.39 | |
| LLaVA-PLLuM-12B-ncJudge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=LLaVA-1.6-Mistral-7B2026.02 | 58.68 | |
| LLaVA-PLLuM-12B-nc-250715Judge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=LLaVA-1.6-Mistral-7B2026.02 | 57.38 | |
| LLaVA-Bielik-11B-v2.6Judge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=LLaVA-1.6-Mistral-7B2026.02 | 55.47 | |
| LLaVA-PLLuM-12B-ncJudge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=LLaVA-1.6-Vicuna-13B2026.02 | 51.4 | |
| LLaVA-PLLuM-12B-nc-250715Judge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=LLaVA-1.6-Vicuna-13B2026.02 | 49.76 | |
| LLaVA-PLLuM-12B-ncJudge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=Pixtral-12B2026.02 | 49.29 | |
| LLaVA-PLLuM-12B-ncJudge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=Qwen2.5-VL-7B2026.02 | 48.75 | |
| LLaVA-Bielik-11B-v2.6Judge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=LLaVA-1.6-Vicuna-13B2026.02 | 48.71 | |
| LLaVA-PLLuM-12B-nc-250715Judge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=Pixtral-12B2026.02 | 47.38 | |
| LLaVA-Bielik-11B-v2.6Judge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=Pixtral-12B2026.02 | 46.72 | |
| LLaVA-PLLuM-12B-nc-250715Judge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=Qwen2.5-VL-7B2026.02 | 46.69 | |
| LLaVA-Bielik-11B-v2.6Judge Model=llava-onevision-qwen2-7b-si-hf, Reference Baseline Model=Qwen2.5-VL-7B2026.02 | 45.7 |