List your company

Mathematical Reasoning on GSM8K (maj@16, pass@16)

97.12maj@16 AccuracyPPC
89.465691.452893.4495.4272May 28, 2026

Evaluation results

MethodLinks
PPCBackbone=Qwen2.5-Math-7B2026.0597.1298.56
Prompt-OnlyBackbone=Qwen2.5-Math-7B2026.0597.0498.18
BaseBackbone=Qwen2.5-Math-7B2026.0596.9798.1
Plan-TuningBackbone=Qwen2.5-Math-7B2026.0596.8298.33
GRPOBackbone=Qwen2.5-Math-7B2026.0596.3698.4
PTA-GRPOBackbone=Qwen3-4B2026.0595.397.57
PPCBackbone=Qwen2.5-7B2026.0595.2298.48
PPCBackbone=Qwen3-4B2026.0595.1598.03
PTA-GRPOBackbone=Qwen2.5-Math-7B2026.0595.1598.03
Plan-TuningBackbone=Qwen3-4B2026.059597.12
GRPOBackbone=Qwen3-4B2026.0594.9297.04
BaseBackbone=Qwen3-4B2026.0594.8496.89
Prompt-OnlyBackbone=Qwen3-4B2026.0594.6297.27
PTA-GRPOBackbone=Qwen2.5-7B2026.0594.6297.27
Prompt-OnlyBackbone=Qwen2.5-7B2026.0594.4797.12
GRPOBackbone=Qwen2.5-7B2026.0594.3197.57
Plan-TuningBackbone=Qwen2.5-7B2026.0594.0997.19
BaseBackbone=Qwen2.5-7B2026.0593.7197.04
PPCBackbone=Llama3.1-8B2026.0592.0498.03
Plan-TuningBackbone=Llama3.1-8B2026.0590.8396.13
PTA-GRPOBackbone=Llama3.1-8B2026.0590.7596.59
GRPOBackbone=Llama3.1-8B2026.0590.5996.36
Prompt-OnlyBackbone=Llama3.1-8B2026.0590.3796.47
BaseBackbone=Llama3.1-8B2026.0589.7695.98