List your company
ResearchFormerly WizWand

Mathematical Reasoning on MathVerse Vision Only

67AccuracyBee-8B
1.68818.64435.652.556Jun 8, 2025Aug 9, 2025Oct 10, 2025Dec 12, 2025Feb 12, 2026Apr 15, 2026Jun 17, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Bee-8BOpenness=fully open, Model Size=8B, Training Strategy=RL2025.1067
Bee-8BOpenness=fully open, Model Size=8B, Training Strategy=SFT2025.1061.9
InternVL3.5Openness=semi-open, Model Size=8B2025.1061.5
Keye-VLOpenness=semi-open, Model Size=8B2025.1059.8
InternVL3.5-2BThinking Mode=true2025.1253.4
MM-Eureka-Qwen-32BActivation Replay=true2025.1152.4
Claude-3.5 SonnetActivation Replay=false2025.1152
MM-Eureka-Qwen-32BActivation Replay=false2025.1150.5
Perception-R1-7BBase Model Family=Qwen2.5-VL2025.0650.1
VL-Rethinker-7BActivation Replay=true2025.1149.2
QvQ-72B-PreviewActivation Replay=false2025.1148.2
MM-Eureka-Qwen-7BActivation Replay=true2025.1147.7
MM-Eureka-7BBase Model Family=Qwen2.5-VL2025.0647.6
Qwen2.5-VL-7B-IT + GRPOBase Model Family=Qwen2.5-VL, Instruction-Tuning=IT, Training Strategy=GRPO2025.0647.1
Vision-R1-7BBase Model Family=Qwen2.5-VL2025.0647
VL-Rethinker-7BActivation Replay=false2025.1147
VLAA-Thinker-7BBase Model Family=Qwen2.5-VL2025.0645.7
MM-Eureka-Qwen-7BActivation Replay=false2025.1145.1
V-ZeroBackbone=Qwen2.5-VL-7B-Instruct, Iteration=12026.0143.9
Gemini-2.0-FlashActivation Replay=false2025.1143.6
SophiaVL-R1-7BBase Model Family=Qwen2.5-VL2025.0643.3
MMR1-Math-v0-7BActivation Replay=true2025.1143.3
Qwen2.5-VLModel Size=7B2026.0642.7
V-ZeroBackbone=Qwen2.5-VL-7B-Instruct, Iteration=22026.0142.6
Qwen2.5-VL-7B-ITBase Model Family=Qwen2.5-VL, Instruction-Tuning=IT2025.0642.2
Supervised GRPOBackbone=Qwen2.5-VL-7B-Instruct2026.0142
R1-Onevision-7BBase Model Family=Qwen2.5-VL2025.0641.9
Qwen2.5-VL-7BActivation Replay=false2025.1141.1
MMR1-Math-v0-7BActivation Replay=false2025.1141.1
GPT-4oActivation Replay=false2025.1140.6
Qwen2.5-VL-7B-Instruct (Base)Backbone=Qwen2.5-VL-7B-Instruct2026.0139.6
OpenVLThinker-7BBase Model Family=Qwen2.5-VL2025.0639.5
Perception-R1-Qwen2-7BBase Model Family=Qwen2-VL2025.0639.2
OpenVLThinker-7BBackbone=Qwen2.5-VL-7B-Instruct2026.0136.8
R1-VL-7BBase Model Family=Qwen2-VL2025.0636.8
SDAR-VLModel Size=8B2026.0636.5
InternVL3.5-2BThinking Mode=false, Evaluation Framework=VLMEvalKit2025.1235.3
Kimi-VL-16BActivation Replay=false2025.1134.1
InternVL3-8BActivation Replay=false2025.1133.9
V-ZeroBackbone=Qwen2.5-VL-3B-Instruct, Iteration=12026.0133.4
V-ZeroBackbone=Qwen2.5-VL-3B-Instruct, Iteration=22026.0132.6
Qwen2-VL-7B-IT + GRPOBase Model Family=Qwen2-VL, Instruction-Tuning=IT, Training Strategy=GRPO2025.0632.4
Qwen2.5-VL-3B-Instruct (Base)Backbone=Qwen2.5-VL-3B-Instruct2026.0132.1
Qwen2-VL-7B-ITBase Model Family=Qwen2-VL, Instruction-Tuning=IT2025.0630.1
GPT-4o-miniActivation Replay=false2025.1130
LaViDaModel Size=8B2026.0627.2
LLaVA OneVisionOpenness=fully open, Model Size=7B2025.1026.2
InternVL3-2B2025.1225.3
PerceptionDLM-BaseModel Size=8B2026.0625.3
Qwen2.5-VLOpenness=semi-open, Model Size=7B2025.1025.1
jina-vlm2025.1223.9
InternVL3Model Size=8B2026.0623.1
Qwen3-VL-2BEvaluation Framework=VLMEvalKit2025.1222.7
LLaDA-VModel Size=8B2026.0620.6
LLaVA-OV-7BActivation Replay=false2025.1117.6
Qwen2.5-VL-3BActivation Replay=false2025.1117.5
Qwen2-VL-2BEvaluation Framework=VLMEvalKit2025.1217.3
MolmoOpenness=fully open, Model Size=7B, Training Strategy=Distilled (-D)2025.104.2