CORE MACHINE LEARNING

SYSTEMS RESEARCH

Latent Execution for Neural Program Synthesis

November 01, 2021

Abstract

Program synthesis from input-output (IO) examples has been a long-standing challenge. While recent works demonstrated limited success on domain-specific languages (DSL), it remains highly challenging to apply them to real-world programming languages, such as C. Due to complicated syntax and token variation, there are three major challenges: (1) unlike many DSLs, programs in languages like C need to compile first and are not executed via interpreters; (2) the program search space grows exponentially when the syntax and semantics of the programming language become more complex; and (3) collecting a large-scale dataset of real-world programs is non-trivial. As a first step to address these challenges, we propose LaSynth and show its efficacy in a restricted-C domain (i.e., C code with tens of tokens, with sequential, branching, loop and simple arithmetic operations but no library call). More specifically, LaSynth learns the latent representation to approximate the execution of partially generated programs, even if they are incomplete in syntax (addressing (1)). The learned execution significantly improves the performance of next token prediction over existing approaches, facilitating search (addressing (2)). Finally, once trained with randomly generated groundtruth programs and their IO pairs, LaSynth can synthesize more concise programs that resemble human-written code. Furthermore, retraining our model with these synthesized programs yields better performance with fewer samples for both Karel and C program synthesis, indicating the promise of leveraging the learned program synthesizer to improve the dataset quality for input-output program synthesis (addressing (3)). When evaluating on whether the program execution outputs match the IO pairs, LaSynth achieves 55.2% accuracy on generating simple C code with tens of tokens including loops and branches, outperforming existing approaches without executors by around 20%.

Download the Paper

AUTHORS

Written by

Xinyun Chen

Dawn Song

Yuandong Tian

Publisher

NeurIPS

Research Topics

Systems Research

Core Machine Learning

Related Publications

August 12, 2024

CORE MACHINE LEARNING

Contrastive Predict-and-Search for Mixed Integer Linear Programs

Arman Zharmagambetov, Yuandong Tian, Aaron Ferber, Bistra Dilkina, Taoan Huang

August 12, 2024

August 09, 2024

CORE MACHINE LEARNING

Benchmarking Attacks on Learning with Errors

Emily Wenger, Eshika Saxena, Mohamed Malhou, Ellie Thieu, Kristin Lauter

August 09, 2024

August 02, 2024

CORE MACHINE LEARNING

GenCO: Generating Diverse Designs with Combinatorial Constraints

Arman Zharmagambetov, Yuandong Tian

August 02, 2024

July 29, 2024

COMPUTER VISION

CORE MACHINE LEARNING

Factorizing Text-to-Video Generation by Explicit Image Conditioning

Rohit Girdhar, Mannat Singh, Andrew Brown, Quentin Duval, Samaneh Azadi, Saketh Rambhatla, Mian Akbar Shah, Xi Yin, Devi Parikh, Ishan Misra

July 29, 2024

Help Us Pioneer The Future of AI

We share our open source frameworks, tools, libraries, and models for everything from research exploration to large-scale production deployment.