List your company

Mathematical Reasoning on Olympiadbench (Average@16)

60.6Average@16Ours (Shorten Prefix)
20.87231.18641.551.814Jan 26, 2026

Evaluation results

MethodLinks
Ours (Shorten Prefix)Backbone=Qwen3-4B (non-think)2026.0160.6
GRPOBackbone=Qwen3-4B (non-think)2026.0160
RTSBackbone=Qwen3-4B (non-think)2026.0159.3
MixedBackbone=Qwen3-4B (non-think)2026.0155.9
Ours (Shorten Prefix)Backbone=Qwen3-8B-Base2026.0152.3
GRPOBackbone=Qwen3-8B-Base2026.0151.5
BaseBackbone=Qwen3-4B (non-think)2026.0151.3
MixedBackbone=Qwen3-8B-Base2026.0148.6
RTSBackbone=Qwen3-8B-Base2026.0148.6
Ours (Shorten Prefix)Backbone=Qwen3-4B-Base2026.0147.7
GRPOBackbone=Qwen3-4B-Base2026.0147.5
RTSBackbone=Qwen3-4B-Base2026.0146.4
MixedBackbone=Qwen3-4B-Base2026.0145.4
BaseBackbone=Qwen3-4B-Base2026.0128.7
BaseBackbone=Qwen3-8B-Base2026.0122.4