List your company

Mathematical Reasoning on Minerva Math (Average@16 accuracy)

40.1Average Accuracy @16HiLL
13.26820.23427.234.166Apr 1, 2026

Evaluation results

MethodLinks
HiLLBackbone=Qwen2.5-7B-Instruct2026.0440.1
Scaf-GRPOBackbone=Qwen2.5-7B-Instruct2026.0439.8
HiLLw/o TWBackbone=Qwen2.5-7B-Instruct2026.0439.8
SAGEBackbone=Qwen2.5-7B-Instruct2026.0439.3
GRPOBackbone=Qwen2.5-7B-Instruct2026.0439.1
LUFFYBackbone=Qwen2.5-7B-Instruct2026.0439
BaseBackbone=Qwen2.5-7B-Instruct2026.0438.1
HiLLBackbone=Llama-3.2-3B-Instruct2026.0421.7
HiLLw/o TWBackbone=Llama-3.2-3B-Instruct2026.0421.6
GRPOBackbone=Llama-3.2-3B-Instruct2026.0420.5
SAGEBackbone=Llama-3.2-3B-Instruct2026.0420.1
Scaf-GRPOBackbone=Llama-3.2-3B-Instruct2026.0419.4
BaseBackbone=Llama-3.2-3B-Instruct2026.0417.8
LUFFYBackbone=Llama-3.2-3B-Instruct2026.0414.3