ResearchBenchmarksCode Generation on MBPP (avg@4)Follow83.1Pass@4SHEAR68.95672.62876.379.972Apr 25, 2026Evaluation ResultsMethodMethodLinksPass@4SHEARBackbone=Qwen2.5-14B-BaseBackbone=Qwen2.5-14B-Base2026.0483.1GRPOBackbone=Qwen2.5-14B-BaseBackbone=Qwen2.5-14B-Base2026.0482.8SHEARBackbone=Qwen2.5-Coder-7BBackbone=Qwen2.5-Coder-7B2026.0482.2GRPOBackbone=Qwen2.5-Coder-7BBackbone=Qwen2.5-Coder-7B2026.0481.8GRPOBackbone=Llama3.1-8B-I...Backbone=Llama3.1-8B-Instruct2026.0470.7SHEARBackbone=Llama3.1-8B-I...Backbone=Llama3.1-8B-Instruct2026.0469.5