List your company

Math Reasoning on MATH 500 (BA @20, BA @40, BA @60)

86.2BA @20LRM-only
67.89672.64877.482.152May 28, 2026

Evaluation results

MethodLinks
LRM-onlySetting=Same-family, Models=Qwen3-1.7B / Qwen3-14B2026.0586.286.286.2
LRM-onlySetting=Cross-family, Models=Qwen3-1.7B / DeepSeek-R1-Distill-Qwen-14B2026.0582.682.682.6
GlimpRouterSetting=Same-family, Models=Qwen3-1.7B / Qwen3-14B2026.05767880.1
RoRoSetting=Same-family, Models=Qwen3-1.7B / Qwen3-14B2026.0575.579.881.9
TRIMSetting=Same-family, Models=Qwen3-1.7B / Qwen3-14B2026.0575.177.579.8
STEERSetting=Same-family, Models=Qwen3-1.7B / Qwen3-14B2026.0574.377.479.9
RoRoSetting=Cross-family, Models=Qwen3-1.7B / DeepSeek-R1-Distill-Qwen-14B2026.0574.278.280.6
SpecReasonSetting=Same-family, Models=Qwen3-1.7B / Qwen3-14B2026.0573.976.979.4
TRIMSetting=Cross-family, Models=Qwen3-1.7B / DeepSeek-R1-Distill-Qwen-14B2026.0573.876.278.5
SpecCoTSetting=Same-family, Models=Qwen3-1.7B / Qwen3-14B2026.0573.477.878.8
GlimpRouterSetting=Cross-family, Models=Qwen3-1.7B / DeepSeek-R1-Distill-Qwen-14B2026.0573.478.879.2
STEERSetting=Cross-family, Models=Qwen3-1.7B / DeepSeek-R1-Distill-Qwen-14B2026.0573.175.878.8
RSDSetting=Same-family, Models=Qwen3-1.7B / Qwen3-14B2026.0572.875.879.2
SpecReasonSetting=Cross-family, Models=Qwen3-1.7B / DeepSeek-R1-Distill-Qwen-14B2026.0572.675.278.4
SpecCoTSetting=Cross-family, Models=Qwen3-1.7B / DeepSeek-R1-Distill-Qwen-14B2026.0572.175.878
RSDSetting=Cross-family, Models=Qwen3-1.7B / DeepSeek-R1-Distill-Qwen-14B2026.0571.574.377.8
RandomSetting=Same-family, Models=Qwen3-1.7B / Qwen3-14B2026.0570.474.276.1
RandomSetting=Cross-family, Models=Qwen3-1.7B / DeepSeek-R1-Distill-Qwen-14B2026.0569.873.575.4
SRM-onlySetting=Same-family, Models=Qwen3-1.7B / Qwen3-14B2026.0568.668.668.6
SRM-onlySetting=Cross-family, Models=Qwen3-1.7B / DeepSeek-R1-Distill-Qwen-14B2026.0568.668.668.6