This method is designed to overcome challenges in a standard autoregressive approach, allowing the model to make targeted, grammatically correct changesThis method is designed to overcome challenges in a standard autoregressive approach, allowing the model to make targeted, grammatically correct changes

Syntactically Valid Code Editing: A Training Methodology for Neural Program Synthesis

Autor: Hackernoon

2025/09/25 00:00

Table of Links

Abstract and 1. Introduction

Background & Related Work
Method

3.1 Sampling Small Mutations

3.2 Policy

3.3 Value Network & Search

3.4 Architecture
Experiments

4.1 Environments

4.2 Baselines

4.3 Ablations
Conclusion, Acknowledgments and Disclosure of Funding, and References

\

Appendix

A. Mutation Algorithm

B. Context-Free Grammars

C. Sketch Simulation

D. Complexity Filtering

E. Tree Path Algorithm

F. Implementation Details

3.4 Architecture

\ We add two additional types of tokens: an token, which serves as a start-of-sentence token for the model; and tokens, which allow the model to reference positions within its context. Given a current image, a target image, and a current tokenized program, we train this transformer model to predict the edit position and the replacement text autoregressively. While making predictions, the decoding is constrained under the grammar. We mask out the prediction logits to only include edit positions that represent nodes in the syntax tree, and only produce replacements that are syntactically valid for the selected edit position.

\ We set σsmall = 2, which means the network is only allowed to produce edits with fewer than two primitives. For training data, we sample an infinite stream of random expressions from the CFG. We choose a random number of noise steps, s ∈ [1, 5], to produce a mutated expression. For some percentage of the examples, ρ, we instead sample a completely random new expression as our mutated expression. We trained for 3 days for the environments we tested on a single Nvidia A6000 GPU.

:::info Authors:

(1) Shreyas Kapur, University of California, Berkeley (srkp@cs.berkeley.edu);

(2) Erik Jenner, University of California, Berkeley (jenner@cs.berkeley.edu);

(3) Stuart Russell, University of California, Berkeley (russell@cs.berkeley.edu).

:::

:::info This paper is available on arxiv under CC BY-SA 4.0 DEED license.

:::

Aviso legal: Los artículos republicados en este sitio provienen de plataformas públicas y se ofrecen únicamente con fines informativos. No reflejan necesariamente la opinión de MEXC. Todos los derechos pertenecen a los autores originales. Si consideras que algún contenido infringe derechos de terceros, comunícate con service@support.mexc.com para solicitar su eliminación. MEXC no garantiza la exactitud, la integridad ni la actualidad del contenido y no se responsabiliza por acciones tomadas en función de la información proporcionada. El contenido no constituye asesoría financiera, legal ni profesional, ni debe interpretarse como recomendación o respaldo por parte de MEXC.

Compartir perspectivas