Understanding by Reconstruction: Reversing the Software Development Process for LLM Pretraining

TL;DR

This paper introduces a novel pre-training paradigm for LLMs that reconstructs the latent reasoning and development trajectories behind static code repositories, significantly improving their complex reasoning and coding abilities.

Contribution

It proposes a framework for synthesizing and utilizing reconstructed development trajectories via multi-agent simulation and optimization, enhancing LLM pretraining.

Findings

Improved performance on long-context understanding benchmarks

Enhanced coding proficiency of Llama-3-8B

Better agentic reasoning capabilities in LLMs

Abstract

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.