TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/Evaluating Large Language Models Trained on Code

Evaluating Large Language Models Trained on Code

Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, Alex Ray, Raul Puri, Gretchen Krueger, Michael Petrov, Heidy Khlaaf, Girish Sastry, Pamela Mishkin, Brooke Chan, Scott Gray, Nick Ryder, Mikhail Pavlov, Alethea Power, Lukasz Kaiser, Mohammad Bavarian, Clemens Winter, Philippe Tillet, Felipe Petroski Such, Dave Cummings, Matthias Plappert, Fotios Chantzis, Elizabeth Barnes, Ariel Herbert-Voss, William Hebgen Guss, Alex Nichol, Alex Paino, Nikolas Tezak, Jie Tang, Igor Babuschkin, Suchir Balaji, Shantanu Jain, William Saunders, Christopher Hesse, Andrew N. Carr, Jan Leike, Josh Achiam, Vedant Misra, Evan Morikawa, Alec Radford, Matthew Knight, Miles Brundage, Mira Murati, Katie Mayer, Peter Welinder, Bob McGrew, Dario Amodei, Sam McCandlish, Ilya Sutskever, Wojciech Zaremba

2021-07-07Multi-task Language UnderstandingCode GenerationLanguage ModellingHumanEval
PaperPDFCode(official)CodeCodeCodeCodeCodeCodeCodeCodeCodeCodeCodeCode

Abstract

We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics.

Results

TaskDatasetMetricValueModel
Transfer LearningBBH-algAverage (%)73.9code-davinci-002 175B (CoT)
Transfer LearningBBH-nlpAverage (%)73.5code-davinci-002 175B (CoT)
Multi-Task LearningBBH-algAverage (%)73.9code-davinci-002 175B (CoT)
Multi-Task LearningBBH-nlpAverage (%)73.5code-davinci-002 175B (CoT)

Related Papers

Visual-Language Model Knowledge Distillation Method for Image Quality Assessment2025-07-21CUDA-L1: Improving CUDA Optimization via Contrastive Reinforcement Learning2025-07-18Towards Formal Verification of LLM-Generated Code from Natural Language Prompts2025-07-17Making Language Model a Hierarchical Classifier and Generator2025-07-17VisionThink: Smart and Efficient Vision Language Model via Reinforcement Learning2025-07-17The Generative Energy Arena (GEA): Incorporating Energy Awareness in Large Language Model (LLM) Human Evaluations2025-07-17Inverse Reinforcement Learning Meets Large Language Model Post-Training: Basics, Advances, and Opportunities2025-07-17MERA Code: A Unified Framework for Evaluating Code Generation Across Tasks2025-07-16