PECC: Problem Extraction and Coding Challenges

Benchmark Model Rank Results
code-generation-on-peccClaude 3 Haiku#1Pass@3: 27.67
code-generation-on-peccGPT-3.5 Turbo#2Pass@3: 23.75
code-generation-on-pecccodechat-bison#3Pass@3: 11.39
code-generation-on-peccchat-bison#4Pass@3: 8.48
code-generation-on-peccMixtral-8x7B-Instruct#5Pass@3: 8.35
code-generation-on-peccPhi-3-mini-128k-instruct#6Pass@3: 7.18
code-generation-on-peccWizardLM-2-7B#7Pass@3: 3.72
code-generation-on-peccLlama-3-8B-Instruct#8Pass@3: 3.1