List your company

Mathematical Reasoning on MinervaMath (Avg@16)

35.3Avg Accuracy @16HEAPA
5.45213.20120.9528.699Jan 30, 2026

Evaluation results

MethodLinks
HEAPABackbone=Qwen3-8B2026.0135.3
HEAPABackbone=Qwen3-4B2026.0132.6
GRPO (w. PS)Backbone=Qwen3-8B2026.0132.4
GRPOBackbone=Qwen3-8B2026.0128.9
GRPO (w. PS)Backbone=Qwen3-4B2026.0128
GRPOBackbone=Qwen3-4B2026.0121
HEAPABackbone=Qwen3-1.7B2026.0116.1
GRPO (w. PS)Backbone=Qwen3-1.7B2026.0115.6
GRPOBackbone=Qwen3-1.7B2026.0114.8
HEAPABackbone=Qwen3-0.6B2026.018.9
GRPO (w. PS)Backbone=Qwen3-0.6B2026.018
GRPOBackbone=Qwen3-0.6B2026.016.6