List your company
ResearchFormerly WizWand

Linguistic correctness of descriptions on XM3600

85.81Preference RateLLaVA-PLLuM-12B-nc
32.71846.501560.28574.0685Feb 15, 2026
Updated 8mo ago

Evaluation results

MethodLinks
LLaVA-PLLuM-12B-ncJudge=Llama-3.3-70B-Instruct, Baseline model=LLaVA-1.6-Mistral-7B2026.0285.81
LLaVA-PLLuM-12B-nc-250715Judge=Llama-3.3-70B-Instruct, Baseline model=LLaVA-1.6-Mistral-7B2026.0284.91
LLaVA-Bielik-11B-v2.6Judge=Llama-3.3-70B-Instruct, Baseline model=LLaVA-1.6-Mistral-7B2026.0282.35
LLaVA-PLLuM-12B-ncJudge=Llama-3.3-70B-Instruct, Baseline model=PaliGemma2-10B2026.0277.53
LLaVA-PLLuM-12B-nc-250715Judge=Llama-3.3-70B-Instruct, Baseline model=PaliGemma2-10B2026.0277.47
LLaVA-Bielik-11B-v2.6Judge=Llama-3.3-70B-Instruct, Baseline model=PaliGemma2-10B2026.0274.1
LLaVA-PLLuM-12B-ncJudge=Llama-3.3-70B-Instruct, Baseline model=LLaVA-1.6-Vicuna-13B2026.0266.71
LLaVA-PLLuM-12B-nc-250715Judge=Llama-3.3-70B-Instruct, Baseline model=LLaVA-1.6-Vicuna-13B2026.0263.64
LLaVA-Bielik-11B-v2.6Judge=Llama-3.3-70B-Instruct, Baseline model=LLaVA-1.6-Vicuna-13B2026.0260.32
LLaVA-PLLuM-12B-ncJudge=Llama-3.3-70B-Instruct, Baseline model=Pixtral-12B2026.0248.33
LLaVA-PLLuM-12B-nc-250715Judge=Llama-3.3-70B-Instruct, Baseline model=Pixtral-12B2026.0243.38
LLaVA-PLLuM-12B-ncJudge=Llama-3.3-70B-Instruct, Baseline model=Qwen2.5-VL-7B2026.0243.15
LLaVA-PLLuM-12B-nc-250715Judge=Llama-3.3-70B-Instruct, Baseline model=Qwen2.5-VL-7B2026.0242.69
LLaVA-Bielik-11B-v2.6Judge=Llama-3.3-70B-Instruct, Baseline model=Pixtral-12B2026.0240.31
LLaVA-Bielik-11B-v2.6Judge=Llama-3.3-70B-Instruct, Baseline model=Qwen2.5-VL-7B2026.0234.76