List your company
ResearchFormerly WizWand

General Visual Question Answering on MMStar

77.8ScoreSeed 1.5-VL
26.89240.108553.32566.5415Feb 19, 2025May 13, 2025Aug 5, 2025Oct 27, 2025Jan 19, 2026Apr 12, 2026Jul 5, 2026
Updated 3mo ago

Evaluation results

MethodLinks
Seed 1.5-VLthinking=true, decoding=greedy2025.0577.8
Gemini 1.5 Prothinking=true, decoding=greedy2025.0577.5
Gemini 2.5 Pro2026.0477.5
Seed 1.5-VLthinking=false, decoding=greedy2025.0576.2
Keye-VLOpenness=semi-open, Model Size=8B2025.1075.5
OpenVLThinkerV2Parameters=8B2026.0473.8
Qwen3-VL GRPORL Strategy=GRPO2026.0471.7
Qwen3-VL GDPORL Strategy=GDPO2026.0471.5
Bee-8BOpenness=fully open, Model Size=8B, Training Strategy=RL2025.1071.4
Qwen2.5-VLNumber of Parameters=72B2025.0270.8
Qwen 2.5-VL 72Bthinking=false, decoding=greedy2025.0570.8
OneThinker-8BParameters=8B2026.0470.6
InternVL2.5Number of Parameters=78B2025.0269.5
InternVL3.5Openness=semi-open, Model Size=8B2025.1069.3
Bee-8BOpenness=fully open, Model Size=8B, Training Strategy=SFT2025.1069
Claude 3.7 Sonnetthinking=true, decoding=sampling2025.0568.8
Qwen3-VL-Instruct-8BParameters=8B2026.0468.5
Qwen2-VLNumber of Parameters=72B2025.0268.3
CPOModel=Qwen3-VL-4B2026.0767.73
OpenAI o1thinking=true, decoding=greedy2025.0567.5
VisionZero2026.0465.8
ARES-7BParameters=7B2026.0465.3
Claude-3.5 Sonnet-0620Model Version=06202025.0265.1
GPT-4othinking=false, decoding=greedy2025.0565.1
GPT-4o2026.0465.1
GPT-4o 0513Model Version=05132025.0264.7
MM-Eureka-7BParameters=7B2026.0464.4
Qwen2.5-VLNumber of Parameters=7B2025.0263.9
Qwen2.5-VLOpenness=semi-open, Model Size=7B2025.1063.9
GRPOModel=Qwen3-VL-4B2026.0763.2
Vision-G12026.0463.1
VL-Rethinker-7BParameters=7B2026.0462.7
OVR-7BParameters=7B2026.0462.7
LoRAModel=Qwen3-VL-4B2026.0761.93
LLaVA OneVisionOpenness=fully open, Model Size=7B2025.1061.7
OpenVLThinker-7BParameters=7B2026.0459.1
BaseModel=Qwen3-VL-4B2026.0758.93
AVLM-2BModel Scale=2B2026.0657.1
MolmoOpenness=fully open, Model Size=7B, Training Strategy=Distilled (-D)2025.1056.1
Qwen2.5-VLNumber of Parameters=3B2025.0255.9
FFTModel=Qwen3-VL-4B2026.0755.27
Qwen2.5-Omni-3BModel Scale=3B2026.0653.82
GRPOFine-tuning setting=GRPO, Backbone=Qwen3-VL-2B2026.0753.47
LoRAFine-tuning setting=LoRA, Backbone=Qwen3-VL-2B2026.0753.27
CPOFine-tuning setting=CPO, Backbone=Qwen3-VL-2B2026.0753.13
BaseFine-tuning setting=Base, Backbone=Qwen3-VL-2B2026.0752.8
GSPOModel=Qwen3-VL-4B2026.0751.2
GSPOFine-tuning setting=GSPO, Backbone=Qwen3-VL-2B2026.0750.53
FFTFine-tuning setting=FFT, Backbone=Qwen3-VL-2B2026.0748.6
Gemma-4-E2B-itModel Scale=4B2026.0628.85