List your company

Mathematical Problem Solving on MATH Hard OOD (test)

96.09AccuracySkillFlow
87.156489.475791.79594.1143May 13, 2026

Evaluation results

MethodLinks
SkillFlowMethod Type=Ours, Backbone Model=Qwen3.5-9B2026.0596.09
FlowSteerMethod Type=Agent+RL, Backbone Model=Qwen3.5-9B2026.0593.75
AgentFlowMethod Type=Agent+RL, Backbone Model=Qwen3.5-9B2026.0592.19
v4-flashMethod Type=Baseline, Backbone Model=v4-flash2026.0591.41
SkillRLMethod Type=Agent+RL, Backbone Model=Qwen3.5-9B2026.0591.41
AFlowMethod Type=Search-based workflow, Backbone Model=Qwen3.5-9B2026.0590.63
Qwen3.5Method Type=Baseline, Backbone Model=Qwen3.5-9B2026.0589.06
Qwen3.5Method Type=SFT, Backbone Model=Qwen3.5-9B2026.0588.28
Qwen3.5Method Type=GRPO, Backbone Model=Qwen3.5-9B2026.0587.5