List your company

Mathematical Reasoning on GSM8K, MATH 500, and TheoremQA

68.14Math ScoreA-CHORD
37.335245.332653.3361.3274Sep 29, 2025

Evaluation results

MethodLinks
A-CHORDBackbone=Qwen3-4B2025.0968.14
CHORDBackbone=Qwen3-4B2025.0967.67
GRPOBackbone=Qwen3-4B2025.0966.52
A-GRPOBackbone=Qwen3-4B2025.0965.12
A-SFT → A-GRPOBackbone=Qwen3-4B2025.0963.65
SFT → A-GRPOBackbone=Qwen3-4B2025.0960.53
Qwen3-4BBackbone=Qwen3-4B2025.0959.47
SFT → GRPOBackbone=Qwen3-4B2025.0959.4
SFTBackbone=Qwen3-4B2025.0956.92
A-SFTBackbone=Qwen3-4B2025.0956.65
A-SFT → A-GRPOBackbone=Qwen3-0.6B2025.0943.72
Qwen3-0.6BBackbone=Qwen3-0.6B2025.0942.29
GRPOBackbone=Qwen3-0.6B2025.0941.49
A-GRPOBackbone=Qwen3-0.6B2025.0941.04
A-SFTBackbone=Qwen3-0.6B2025.0940.14
SFT → A-GRPOBackbone=Qwen3-0.6B2025.0939.95
SFT → GRPOBackbone=Qwen3-0.6B2025.0939.57
SFTBackbone=Qwen3-0.6B2025.0939.56
A-CHORDBackbone=Qwen3-0.6B2025.0939.45
CHORDBackbone=Qwen3-0.6B2025.0938.52