List your company

Step Verification on VisualProcessBench (test)

71.32Macro-F1 (DynaMath)Gemini 2.5 Pro
48.533654.449360.36566.2807Mar 17, 2026Mar 26, 2026Apr 5, 2026Apr 15, 2026Apr 25, 2026May 5, 2026May 15, 2026

Evaluation results

MethodLinks
Gemini 2.5 ProPrompting Condition=structured constraints (Yes)2026.0371.3264.4269.7865.2672.4368.64—
EVPV-PRM2026.0369.5768.8667.0965.2769.1167.46—
doubao-seed-1.6-visionPrompting Condition=structured constraints (Yes)2026.0368.6661.8665.5762.5164.6264.91—
Gemini 2.5 ProPrompting Condition=original prompt (No)2026.0368.4763.3468.2665.1569.4867.13—
Qwen3-VL-235B-A22B-instructPrompting Condition=structured constraints (Yes)2026.0368.4361.5465.9764.5464.1665.45—
doubao-seed-1.6-visionPrompting Condition=original prompt (No)2026.0366.1959.4763.1261.0762.7462.77—
TIM-PRM-8B2026.0365.958.361.958.363.961.7—
BETAPRMBackbone=Qwen2.5-VL-7B2026.0563.7559.4962.9162.1966.69—62.91
BETAPRMBackbone=InternVL2.5-8B2026.056360.4860.760.4359.98—60.97
Standard PRMBackbone=InternVL3-14B2026.0562.9363.1261.9359.963.79—61.9
Standard PRMBackbone=InternVL2.5-8B2026.0562.9162.0560.4760.7864.38—61.54
Standard PRMBackbone=Qwen2.5-VL-7B2026.0562.8861.4461.2562.1765.55—62.23
BETAPRMBackbone=InternVL3-8B2026.0562.7162.4262.1759.6564.23—61.85
VisualPRM-8B2026.0362.758.56162.161.862—
Qwen3-VL-30B-A3B-instructPrompting Condition=structured constraints (Yes)2026.0362.275959.6856.4959.559.26—
BETAPRMBackbone=InternVL3-14B2026.0561.6759.1862.7460.9464.59—61.9
Standard PRMBackbone=InternVL3-8B2026.0561.6358.7360.4160.263.23—60.69
qwen2.5-vl-72b-instructPrompting Condition=structured constraints (Yes)2026.0361.4360.2559.8559.1259.7259.99—
Qwen3-VL-30B-A3B-instructPrompting Condition=original prompt (No)2026.0358.9561.2957.3757.4958.7658.22—
QWEN-VL-PRM-7B2026.0358.355.858.855.759.858.6—
gpt-4o-miniPrompting Condition=structured constraints (Yes)2026.0358.1353.254.0952.0754.6254.29—
Qwen3-VL-235B-A22B-instructPrompting Condition=original prompt (No)2026.0357.6358.7358.0859.5958.7658.51—
qwen2.5-vl-72b-instructPrompting Condition=original prompt (No)2026.0356.9959.4356.4358.0955.7257.19—
gpt-4o-miniPrompting Condition=original prompt (No)2026.0356.5754.0852.5351.4256.7453.57—
Base (w/o training)Backbone=InternVL2.5-8B2026.0554.8550.2152.0452.449.95—52.28
Base (w/o training)Backbone=InternVL3-14B2026.0550.1948.8650.9747.6147.34—49.4
Base (w/o training)Backbone=Qwen2.5-VL-7B2026.0549.6249.8549.5850.2248.51—49.68
Base (w/o training)Backbone=InternVL3-8B2026.0549.4147.0148.9147.2149—48.39