List your company

Mathematical Reasoning on AIME 2025 (Average@16)

29.2Average@16Ours (Shorten Prefix)
2.2649.25716.2523.243Jan 26, 2026

Evaluation results

MethodLinks
Ours (Shorten Prefix)Backbone=Qwen3-4B (non-think)2026.0129.2
GRPOBackbone=Qwen3-4B (non-think)2026.0126
RTSBackbone=Qwen3-4B (non-think)2026.0122.3
MixedBackbone=Qwen3-4B (non-think)2026.0119.6
Ours (Shorten Prefix)Backbone=Qwen3-8B-Base2026.0118.3
BaseBackbone=Qwen3-4B (non-think)2026.0118.1
GRPOBackbone=Qwen3-8B-Base2026.0117.2
Ours (Shorten Prefix)Backbone=Qwen3-4B-Base2026.0116.6
RTSBackbone=Qwen3-8B-Base2026.0116.3
MixedBackbone=Qwen3-8B-Base2026.0115
RTSBackbone=Qwen3-4B-Base2026.0114
MixedBackbone=Qwen3-4B-Base2026.0112.3
GRPOBackbone=Qwen3-4B-Base2026.0111.9
BaseBackbone=Qwen3-4B-Base2026.015.2
BaseBackbone=Qwen3-8B-Base2026.013.3