List your company

Visual Question Answering on ScienceQA (val)

82.9AccuracySNs (mean)
47.0256.33565.6574.965Jun 11, 2025Jul 26, 2025Sep 10, 2025Oct 25, 2025Dec 10, 2025Jan 24, 2026Mar 11, 2026

Evaluation results

MethodLinks
SNs (mean)Backbone=Qwen3-VL-4b-Instruct2026.0382.98271.778.374.8
SNs (maj. voting)Backbone=Qwen3-VL-4b-Instruct2026.0382.48272.973.473.2
Qwen3-VL-4b-Instruct2026.0381.1—68.378.773.1
Uni-DPOModel=Qwen2-VL-2B2025.0668.1————
SimPOModel=Qwen2-VL-2B2025.0666.81————
BaselineModel=Qwen2-VL-2B2025.0665.57————
DPOModel=Qwen2-VL-2B2025.0663————
SNs (mean)Backbone=LLaVA-v1.5-7b2026.0357.46342.484.556.5
LLaVA-v1.5-7b2026.0353.5—40.185.954.7
SNs (maj. voting)Backbone=LLaVA-v1.5-7b2026.0348.46338.394.954.6