List your company

Math Reasoning on MATH 500 (Average@16 accuracy)

81.8Average Accuracy@16HiLL
37.18448.76760.3571.933Apr 1, 2026

Evaluation results

MethodLinks
HiLLBackbone=Qwen2.5-7B-Instruct2026.0481.8
LUFFYBackbone=Qwen2.5-7B-Instruct2026.0481.3
HiLLw/o TWBackbone=Qwen2.5-7B-Instruct2026.0480.5
SAGEBackbone=Qwen2.5-7B-Instruct2026.0480
GRPOBackbone=Qwen2.5-7B-Instruct2026.0479.2
Scaf-GRPOBackbone=Qwen2.5-7B-Instruct2026.0478
BaseBackbone=Qwen2.5-7B-Instruct2026.0475.7
HiLLBackbone=Llama-3.2-3B-Instruct2026.0457.7
HiLLw/o TWBackbone=Llama-3.2-3B-Instruct2026.0456.7
SAGEBackbone=Llama-3.2-3B-Instruct2026.0456.3
GRPOBackbone=Llama-3.2-3B-Instruct2026.0452.1
Scaf-GRPOBackbone=Llama-3.2-3B-Instruct2026.0451.7
BaseBackbone=Llama-3.2-3B-Instruct2026.0444.7
LUFFYBackbone=Llama-3.2-3B-Instruct2026.0438.9