NoviCode: Generating Programs from Natural Language Utterances by Novices

TL;DR

NoviCode introduces a new task of translating natural language descriptions by non-programmers into executable code, highlighting the challenges and proposing a hierarchical alignment approach to improve model performance.

Contribution

The paper presents NoviCode, a novel NL programming task and benchmark, demonstrating that hierarchical alignment significantly enhances code generation from non-technical instructions.

Findings

NoviCode is a challenging new task in code synthesis.

Hierarchical alignment improves LLM performance on this task.

Generating complex code from non-technical descriptions exceeds current Text-to-Code models.

Abstract

Current Text-to-Code models demonstrate impressive capabilities in generating executable code from natural language snippets. However, current studies focus on technical instructions and programmer-oriented language, and it is an open question whether these models can effectively translate natural language descriptions given by non-technical users and express complex goals, to an executable program that contains an intricate flow - composed of API access and control structures as loops, conditions, and sequences. To unlock the challenge of generating a complete program from a plain non-technical description we present NoviCode, a novel NL Programming task, which takes as input an API and a natural language description by a novice non-programmer and provides an executable program as output. To assess the efficacy of models on this task, we provide a novel benchmark accompanied by test suites wherein the generated program code is assessed not according to their form, but according to their functional execution. Our experiments show that, first, NoviCode is indeed a challenging task in the code synthesis domain, and that generating complex code from non-technical instructions goes beyond the current Text-to-Code paradigm. Second, we show that a novel approach wherein we align the NL utterances with the compositional hierarchical structure of the code, greatly enhances the performance of LLMs on this task, compared with the end-to-end Text-to-Code counterparts.