Textbooks Are All You Need II: phi-1.5 technical report

Benchmark Model Rank Results
code-generation-on-mbppphi-1.5-web 1.3B#68Accuracy: 43.5
common-sense-reasoning-on-arc-challengephi-1.5-web 1.3B (zero-shot)#29Accuracy: 44.9
common-sense-reasoning-on-arc-easyphi-1.5-web 1.3B (0-shot)#18Accuracy: 76.1
common-sense-reasoning-on-winograndephi-1.5-web 1.3B (zero-shot)#28Accuracy: 74.0
multi-task-language-understanding-on-mmluphi-1.5-web 1.3B#29Average (%): 37.9
question-answering-on-piqaphi-1.5-web (1.3B)#42Accuracy: 77
question-answering-on-social-iqaphi-1.5-web 1.3B (zero-shot)#16Accuracy: 53.0
question-answering-on-social-iqaphi-1.5 1.3B (zero-shot)#17Accuracy: 52.6