List your company

Mathematical Reasoning on GPQA (Avg@16)

50Avg@16GRPO (w. PS)
30.34435.44740.5545.653Jan 30, 2026

Evaluation results

MethodLinks
GRPO (w. PS)Backbone=Qwen3-8B2026.0150
HEAPABackbone=Qwen3-8B2026.0149.3
GRPOBackbone=Qwen3-8B2026.0145.6
HEAPABackbone=Qwen3-4B2026.0144.3
GRPO (w. PS)Backbone=Qwen3-4B2026.0143.3
GRPOBackbone=Qwen3-4B2026.0141.8
HEAPABackbone=Qwen3-1.7B2026.0133.3
GRPO (w. PS)Backbone=Qwen3-1.7B2026.0132.9
GRPOBackbone=Qwen3-1.7B2026.0132.3
GRPO (w. PS)Backbone=Qwen3-0.6B2026.0131.8
HEAPABackbone=Qwen3-0.6B2026.0131.6
GRPOBackbone=Qwen3-0.6B2026.0131.1