List your company

Mathematical Reasoning on AIME 2024 (Average@16)

31.8Average@16Ours (Shorten Prefix)
5.69612.47319.2526.027Jan 26, 2026

Evaluation results

MethodLinks
Ours (Shorten Prefix)Backbone=Qwen3-4B (non-think)2026.0131.8
RTSBackbone=Qwen3-4B (non-think)2026.0131.2
GRPOBackbone=Qwen3-4B (non-think)2026.0130.4
Ours (Shorten Prefix)Backbone=Qwen3-8B-Base2026.0122.7
GRPOBackbone=Qwen3-8B-Base2026.0122.3
MixedBackbone=Qwen3-4B (non-think)2026.0120.6
BaseBackbone=Qwen3-4B (non-think)2026.0119.5
Ours (Shorten Prefix)Backbone=Qwen3-4B-Base2026.0119.4
GRPOBackbone=Qwen3-4B-Base2026.0118.3
RTSBackbone=Qwen3-4B-Base2026.0117.5
MixedBackbone=Qwen3-8B-Base2026.0116.7
RTSBackbone=Qwen3-8B-Base2026.0115
MixedBackbone=Qwen3-4B-Base2026.0114.9
BaseBackbone=Qwen3-4B-Base2026.017.7
BaseBackbone=Qwen3-8B-Base2026.016.7