List your company

Mathematical Reasoning on SVAMP (Acc, SL, SE)

96.7AccuracyL1
85.15688.15391.1594.147Mar 23, 2026Apr 1, 2026Apr 11, 2026Apr 20, 2026Apr 30, 2026May 9, 2026May 19, 2026

Evaluation results

MethodLinks
L12026.0396.71,1388.5
Cut1024max tokens=10242026.0396.671,8725.16
ROM_CSCCounterfactual Self-Correction=true2026.0395.890610.57
Vanilla2026.03951,8745.07
EAT2026.03951,4806.42
Cut2048max tokens=20482026.03951,8725.07
ROM2026.0392.51,2417.45
Cut512max tokens=5122026.0386.671,8724.63
CESBackbone=DeepSeek-R1-Distill-1.5B2026.0586.4620—
DAPOBackbone=DeepSeek-R1-Distill-1.5B2026.0585.6618—