CodeCompose is an AI-assisted code authoring tool powered by large language models (LLMs) that provides inline suggestions to 10’s of thousands of developers at Meta. In this paper, we present how we scaled the product from displaying single-line suggestions to multi-line suggestions. This evolution required us to overcome several unique challenges in improving the usability of these suggestions for developers.
First, we discuss how multi-line suggestions can have a “jarring” effect, as the LLM’s suggestions constantly move around the developer’s existing code, which would otherwise result in decreased productivity and satisfaction.
Second, multi-line suggestions take significantly longer to generate; hence we present several innovative investments we made to reduce the perceived latency for users. These model-hosting optimizations sped up multi-line suggestion latency by 2.5x.
Finally, we conduct experiments on 10’s of thousands of engineers to understand how multi-line suggestions impact the user experience and contrast this with single-line suggestions. Our experiments reveal that (i) multi-line suggestions account for 42% of total characters accepted (despite only accounting for 16% for displayed suggestions) (ii) multi-line suggestions almost doubled the percentage of keystrokes saved for users from 9% to 17%. Multi-line CodeCompose has been rolled out to all engineers at Meta, and less than 1% of engineers have opted out of multi-line suggestions.
Wed 17 JulDisplayed time zone: Brasilia, Distrito Federal, Brazil change
16:00 - 18:00 | AI4SE 2Industry Papers / Research Papers at Pitomba Chair(s): Jingyue Li Norwegian University of Science and Technology (NTNU) | ||
16:00 18mTalk | MonitorAssistant: Simplifying Cloud Service Monitoring via Large Language Models Industry Papers Zhaoyang Yu Tsinghua University, Minghua Ma Microsoft Research, Chaoyun Zhang Microsoft, Si Qin Microsoft Research, Yu Kang Microsoft Research, Chetan Bansal Microsoft Research, Saravan Rajmohan Microsoft, Yingnong Dang Microsoft Azure, Changhua Pei Computer Network Information Center at Chinese Academy of Sciences, Dan Pei Tsinghua University, Qingwei Lin Microsoft, Dongmei Zhang Microsoft Research | ||
16:18 18mTalk | Code-Aware Prompting: A study of Coverage guided Test Generation in Regression Setting using LLM Research Papers Gabriel Ryan Columbia University, Siddhartha Jain AWS AI Labs, Mingyue Shang AWS AI Labs, Shiqi Wang AWS AI Labs, Xiaofei Ma AWS AI Labs, Murali Krishna Ramanathan AWS AI Labs, Baishakhi Ray Columbia University, New York; AWS AI Lab | ||
16:36 18mTalk | A Machine Learning-Based Error Mitigation Approach for Reliable Software Development on IBM’s Quantum Computers Industry Papers Asmar Muqeet Simula Research Laboratory and University of Oslo, Shaukat Ali Simula Research Laboratory and Oslo Metropolitan University, Tao Yue Beihang University, Paolo Arcaini National Institute of Informatics
| ||
16:54 18mTalk | Multi-line AI-assisted Code Authoring Industry Papers Omer Dunay Meta Platforms, Inc., Daniel Cheng Meta Platforms Inc., Adam Tait Meta Platforms, Inc., Parth Thakkar Meta Platforms, Inc., Peter C Rigby Meta / Concordia University, Andy Chiu Meta Platforms, Inc., Imad Ahmad Meta Platforms, Inc., Arun Ganesan Meta Platforms, Inc., Chandra Sekhar Maddila Meta Platforms, Inc., Vijayaraghavan Murali Meta Platforms Inc., Ali Tayyebi Meta Platforms Inc., Nachiappan Nagappan Meta Platforms, Inc. | ||
17:12 18mTalk | Combating Missed Recalls in E-commerce Search: a CoT-prompting Testing Approach Industry Papers Shengnan Wu School of Computer Science, Fudan University, Yongxiang Hu Fudan University, Yingchuan Wang School of Computer Science, Fudan University, Jiazhen Gu The Chinese University of Hong Kong, Jin Meng Meituan Inc., Liujie Fan Meituan Inc., Zhongshi Luan Meituan Inc., Xin Wang Fudan University, Yangfan Zhou Fudan University Pre-print | ||
17:30 18mTalk | Automated Unit Test Improvement using Large Language Models at Meta Industry Papers Mark Harman Meta Platforms, Inc. and UCL, Jubin Chheda Meta platforms, Anastasia Finogenova Meta platforms, Inna Harper Meta, Alexandru Marginean Meta platforms, Shubho Sengupta Meta platforms, Eddy Wang Meta platforms, Nadia Alshahwan Meta Platforms, Beliz Gokkaya Meta Platforms | ||