List your company

Mathematical Reasoning on SVAMP (Acc, # L, Acc./# L)

56.8AccuracyRuPLaR-Temp
20.29629.77339.2548.727May 10, 2026

Evaluation results

MethodLinks
RuPLaR-Temp2026.0556.8—37.9
RuPLaR-Gumbel2026.0554.81.5635.1
RuPLaR-Mix2026.0554.31.5934.2
RuPLaRPrior Strength (alpha_kl)=0.32026.0552.51.4336.7
RuPLaRPrior Strength (alpha_kl)=2.02026.0551.41.5732.7
RuPLaRLoss Component=w/o Problem-Thought2026.05511.5632.7
ReGuLaR2026.0550.12.0224.8
RuPLaRLoss Component=w/o KL-Focus2026.0548.91.630.6
RuPLaRPrior Type=Random2026.0548.61.4234.2
CoLaR-52026.0548.42.9516.4
RuPLaRPrior Type=Uniform2026.0547.31.5630.3
RuPLaRPrior Type=Learned2026.0545.21.6627.2
COCONUT2026.0540.766.78
iCoT2026.0536.40—
CODI2026.0521.763.62