List your company
ResearchFormerly WizWand

Multimodal Mathematical Reasoning on GEOQA-8k (test)

65AccuracyD2Ijus
27.45637.20346.9556.697Jul 9, 2025Aug 11, 2025Sep 14, 2025Oct 18, 2025Nov 21, 2025Dec 25, 2025Jan 28, 2026
Updated 5mo ago

Evaluation results

MethodLinks
D2IjusReasoning Strategy=D2I, Variant=jus, Backbone=Qwen2.5-VL-7B2025.0765
D2IlocReasoning Strategy=D2I, Variant=loc, Backbone=Qwen2.5-VL-7B2025.0760.6
D2IparReasoning Strategy=D2I, Variant=par, Backbone=Qwen2.5-VL-7B2025.0760.5
DGPOBackbone=Qwen2.5-VL-3B-Instruct2026.0159.95
GPGBackbone=Qwen2.5-VL-3B-Instruct2026.0159.02
DAPOBackbone=Qwen2.5-VL-3B-Instruct2026.0159.02
GRPO-ADBackbone=Qwen2.5-VL-3B-Instruct2026.0158.09
Dr.GRPOBackbone=Qwen2.5-VL-3B-Instruct2026.0157.96
GRPOBackbone=Qwen2.5-VL-3B-Instruct2026.0157.43
GSPOBackbone=Qwen2.5-VL-3B-Instruct2026.0157.16
Qwen2.5-VL-7B w/ GRPOReasoning Mode=Deliberate, Backbone=Qwen2.5-VL-7B2025.0754.9
D2DlocReasoning Strategy=D2D, Variant=loc, Backbone=Qwen2.5-VL-7B2025.0754.5
Qwen2.5-VL-7B w/ GRPO†Reasoning Mode=Intuitive, Backbone=Qwen2.5-VL-7B2025.0753.1
D2DparReasoning Strategy=D2D, Variant=par, Backbone=Qwen2.5-VL-7B2025.0752.1
Qwen2.5-VL-7B*Backbone=Qwen2.5-VL-7B2025.0746.6
Base ModelBackbone=Qwen2.5-VL-3B-Instruct2026.0139.79
D2DjusReasoning Strategy=D2D, Variant=jus, Backbone=Qwen2.5-VL-7B2025.0728.9