List your company

Mathematical Reasoning on GSM8K (Acc, Tok)

86.8AccuracyInfoDensity
77.4479.8782.384.73Mar 18, 2026

Evaluation results

MethodLinks
InfoDensityBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.0386.81,352
DSBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.0386.51,551
GRPO-AccBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.0386.41,757
GRPO-LPBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.0386.41,560
OriginalBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.0385.42,558
PEARBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.0385.11,046
GRPO-AccBackbone=Qwen3-0.6B2026.0382.92,310
PEARBackbone=Qwen3-0.6B2026.0380.21,190
OriginalBackbone=Qwen3-0.6B2026.0378.32,013
InfoDensityBackbone=Qwen3-0.6B2026.0378.3865
DSBackbone=Qwen3-0.6B2026.03781,278
GRPO-LPBackbone=Qwen3-0.6B2026.0377.81,212