Hierarchical Neural Program Synthesis
This addresses a key bottleneck in program synthesis for applications requiring complex code generation, though it is incremental in its approach.
The paper tackles the scalability problem in program synthesis by introducing a hierarchical framework that composes programs from task embeddings, enabling synthesis of significantly longer and more complex programs than prior methods.
Program synthesis aims to automatically construct human-readable programs that satisfy given task specifications, such as input/output pairs or demonstrations. Recent works have demonstrated encouraging results in a variety of domains, such as string transformation, tensor manipulation, and describing behaviors of embodied agents. Most existing program synthesis methods are designed to synthesize programs from scratch, generating a program token by token, line by line. This fundamentally prevents these methods from scaling up to synthesize programs that are longer or more complex. In this work, we present a scalable program synthesis framework that instead synthesizes a program by hierarchically composing programs. Specifically, we first learn a task embedding space and a program decoder that can decode a task embedding into a program. Then, we train a high-level module to comprehend the task specification (e.g., input/output pairs or demonstrations) from long programs and produce a sequence of task embeddings, which are then decoded by the program decoder and composed to yield the synthesized program. We extensively evaluate our proposed framework in a string transformation domain with input/output pairs. The experimental results demonstrate that the proposed framework can synthesize programs that are significantly longer and more complex than the programs considered in prior program synthesis works. Website at https://thoughtp0lice.github.io/hnps_web/