Share your thoughts, 1 month free Claude Pro on usSee more

Multi-modal Reasoning on MathVista (Accuracy)

79.2Accuracy

AutoNPO

Updated 23d ago

Evaluation Results

Method	Links
AutoNPO 2026.04		79.2
RLEP 2026.04		78.5
ExGRPO 2026.04		77.3
NPO 2026.04		76.6
NPO 2026.04		76.3
GRPO 2026.04		76.2
Qwen3-VL-8B-Instruct 2026.04		73.8
LUFFY 2026.04		73.8
Self-Instruct + Solver Feedback 2025.11		70.3
CoT Cold-Start + Solver Feedback 2025.11		70.1
Self-Instruct 2025.11		69.8
Self-Instruct + CoT-Self-Instruct 2025.11		69.3
Self-Instruct + R-Zero 2025.11		69.3
CoT Cold-Start 2025.11		69.1
Seed Set 2025.11		69
Base Model 2025.11		68.4