List your company
ResearchFormerly WizWand

Multimodal Reasoning on ScienceQA (test)

95.93AccuracyFineViT-VL
52.426863.720975.01586.3091Mar 17, 2026Apr 3, 2026Apr 20, 2026May 7, 2026May 24, 2026Jun 10, 2026Jun 28, 2026
Updated 4mo ago

Evaluation results

MethodLinks
FineViT-VLLanguage Backbone=Qwen3 1.7B LLM, Evaluation Framework=VLMEvalKit [69]2026.0395.93
Aquila-VLLanguage Backbone=Qwen2.5 1.5B LLM, Evaluation Framework=VLMEvalKit [69]2026.0395.33
Intern3.5-VLLanguage Backbone=Qwen3 1.7B LLM, Evaluation Framework=VLMEvalKit [69]2026.0391.17
CIRCLESSize=7B, Model Family=Qwen2.5-VL2026.0388.35
RICESSize=7B, Model Family=Qwen2.5-VL2026.0388.05
MUIERSize=7B, Model Family=Qwen2.5-VL2026.0387.9
Qwen3-VLLanguage Backbone=Qwen3 1.7B LLM, Evaluation Framework=VLMEvalKit [69]2026.0386.47
RandomSize=7B, Model Family=Qwen2.5-VL2026.0385.18
MoD-Multi RoundBackbone=Qwen2.5VL-3b-Instruct, Protocol=multi-turn dialectical reasoning2026.0681.71
Multi-Agent DebateBackbone=Qwen2.5VL-3b-Instruct2026.0681.62
Self-CorrectionBackbone=Qwen2.5VL-3b-Instruct2026.0681.57
Qwen2.5VL-3b-Instruct2026.0681.4
CIRCLESSize=3B, Model Family=Qwen2.5-VL2026.0380.71
MUIERSize=3B, Model Family=Qwen2.5-VL2026.0380.66
MoD-Single RoundBackbone=Qwen2.5VL-3b-Instruct, Protocol=single-round inference2026.0680.66
MoE-LoRABackbone=Qwen2.5VL-3b-Instruct2026.0680.45
RICESSize=3B, Model Family=Qwen2.5-VL2026.0380.12
RandomSize=3B, Model Family=Qwen2.5-VL2026.0378.33
Gemma-3-4B2026.0677.1
MoD-Multi RoundBackbone=LLaVA-v1.6-13b, Protocol=multi-turn dialectical reasoning2026.0675.21
MoE-LoRABackbone=LLaVA-v1.6-13b2026.0674.51
MoD-Single RoundBackbone=LLaVA-v1.6-13b, Protocol=single-round inference2026.0674.51
Multi-Agent DebateBackbone=LLaVA-v1.6-13b2026.0674.31
Self-CorrectionBackbone=LLaVA-v1.6-13b2026.0674.16
LLaVA-v1.6-13b2026.0673.62
LLaVA-Next-Mistral-7B2026.0673
ShareGPT4V-13B2026.0672.6
Qwen-VL-Chat2026.0668.8
MoE-LLaVA-2.7Bx42026.0668.5
MoE-LLaVA-1.8Bx42026.0663.1
InstructBLIP-7B2026.0654.1