Saved in:
Bibliographic Details
Main Authors: Zeng, Zhiyuan, Zhang, Yichi, Shan, Yong, Hua, Kai, Fang, Siyuan, Liu, Zhaiyu, Liu, Jiaheng, Wang, Haozhe, Zheng, Yining, Ding, Ming, Shen, Ke, Zhang, Ge, Huang, Wenhao, Qiu, Xipeng
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.11103
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914409148841984
author Zeng, Zhiyuan
Zhang, Yichi
Shan, Yong
Hua, Kai
Fang, Siyuan
Liu, Zhaiyu
Liu, Jiaheng
Wang, Haozhe
Zheng, Yining
Ding, Ming
Shen, Ke
Zhang, Ge
Huang, Wenhao
Qiu, Xipeng
author_facet Zeng, Zhiyuan
Zhang, Yichi
Shan, Yong
Hua, Kai
Fang, Siyuan
Liu, Zhaiyu
Liu, Jiaheng
Wang, Haozhe
Zheng, Yining
Ding, Ming
Shen, Ke
Zhang, Ge
Huang, Wenhao
Qiu, Xipeng
contents While Large Language Models (LLMs) have achieved remarkable success in code generation, they often struggle with the deep, long-horizon reasoning required for complex software engineering. We attribute this limitation to the nature of standard pre-training data: static software repositories represent only the terminal state of an intricate intellectual process, abstracting away the intermediate planning, debugging, and iterative refinement. To bridge this gap, we propose a novel paradigm: understanding via reconstruction. We hypothesize that reverse-engineering the latent agentic trajectories -- the planning, reasoning, and debugging steps -- behind static repositories provides a far richer supervision signal than raw code alone. To operationalize this, we introduce a framework that synthesizes these trajectories using a multi-agent simulation. This process is grounded in the structural realities of the source repositories (e.g., dependency graphs and file hierarchies) to ensure fidelity. Furthermore, to guarantee the logical rigor of the synthetic data, we employ a search-based optimization technique that iteratively refines the Chain-of-Thought (CoT) reasoning to maximize the likelihood of the ground-truth code. Empirical results demonstrate that continuous pre-training on these reconstructed trajectories significantly enhances Llama-3-8B's performance across diverse benchmarks, including long-context understanding, coding proficiency, and agentic capabilities.
format Preprint
id arxiv_https___arxiv_org_abs_2603_11103
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Understanding by Reconstruction: Reversing the Software Development Process for LLM Pretraining
Zeng, Zhiyuan
Zhang, Yichi
Shan, Yong
Hua, Kai
Fang, Siyuan
Liu, Zhaiyu
Liu, Jiaheng
Wang, Haozhe
Zheng, Yining
Ding, Ming
Shen, Ke
Zhang, Ge
Huang, Wenhao
Qiu, Xipeng
Software Engineering
While Large Language Models (LLMs) have achieved remarkable success in code generation, they often struggle with the deep, long-horizon reasoning required for complex software engineering. We attribute this limitation to the nature of standard pre-training data: static software repositories represent only the terminal state of an intricate intellectual process, abstracting away the intermediate planning, debugging, and iterative refinement. To bridge this gap, we propose a novel paradigm: understanding via reconstruction. We hypothesize that reverse-engineering the latent agentic trajectories -- the planning, reasoning, and debugging steps -- behind static repositories provides a far richer supervision signal than raw code alone. To operationalize this, we introduce a framework that synthesizes these trajectories using a multi-agent simulation. This process is grounded in the structural realities of the source repositories (e.g., dependency graphs and file hierarchies) to ensure fidelity. Furthermore, to guarantee the logical rigor of the synthetic data, we employ a search-based optimization technique that iteratively refines the Chain-of-Thought (CoT) reasoning to maximize the likelihood of the ground-truth code. Empirical results demonstrate that continuous pre-training on these reconstructed trajectories significantly enhances Llama-3-8B's performance across diverse benchmarks, including long-context understanding, coding proficiency, and agentic capabilities.
title Understanding by Reconstruction: Reversing the Software Development Process for LLM Pretraining
topic Software Engineering
url https://arxiv.org/abs/2603.11103