List your company

Multimodal Math Reasoning on MathVista In-domain

65.3Overall AccuracyADPO
57.60459.60261.663.598Jan 4, 2026

Evaluation results

MethodLinks
ADPOSample size=122026.0165.371.759.8
ADPOSample size=82026.016572.258.9
ADPOSample size=42026.0164.871.359.3
MM-VerifierSample size=122026.0164.170.458.7
R1-VL-7BSample size=12026.0163.5——
GRPOSample size=42026.0163.469.858
GRPOSample size=122026.0163.470.757.2
GRPOSample size=82026.0162.970.456.5
MM-VerifierSample size=82026.0162.568.557.4
ADPOSample size=12026.0162.468.757
GRPOSample size=12026.0162.269.855.7
Base (Qwen2-VL-7B)Sample size=122026.0160.767.455
Base (Qwen2-VL-7B)Sample size=82026.0160.16853.3
MM-VerifierSample size=42026.0159.86753.7
Base (Qwen2-VL-7B)Sample size=42026.0158.265.751.9
Base (Qwen2-VL-7B)Sample size=12026.0157.968.948.5