Multi-Location Software Model Completion

TL;DR

The paper tackles the challenge of coordinating multi-location edits in large, evolving software models by introducing Next-Focus, a global embedding-based predictor that recommends additional change locations given an initial anchor edit. Built on an attention-based neural network and history-based co-change signals, Next-Focus iteratively supports multi-location model completion by ranking candidate focus nodes using learned patterns from historical evolution data. Evaluation on real-world modeling histories shows Next-Focus achieving an average Precision@$k$ of about $0.98$ for $k \le 10$, significantly outperforming baselines such as random, semantic similarity, and historical co-change frequency, and it remains effective for changes spread across large parts of the model. When combined with a single-location completion method, Next-Focus enables robust iterative multi-location completion, achieving notable gains in next-focus accuracy and structural correctness. The work demonstrates the practical potential of cross-project knowledge transfer and provides datasets and code to facilitate reproducibility and further research in AI-assisted software modeling.

Abstract

In model-driven engineering and beyond, software models are key development artifacts. In practice, they often grow to substantial size and complexity, undergoing thousands of modifications over time due to evolution, refactoring, and maintenance. The rise of AI has sparked interest in how software modeling activities can be automated. Recently, LLM-based approaches for software model completion have been proposed, however, the state of the art supports only single-location model completion by predicting changes at a specific location. Going beyond, we aim to bridge the gap toward handling coordinated changes that span multiple locations across large, complex models. Specifically, we propose a novel global embedding-based next focus predictor, NextFocus, which is capable of multi-location model completion for the first time. The predictor consists of a neural network with an attention mechanism that is trained on historical software model evolution data. Starting from an existing change, it predicts further model elements to change, potentially spanning multiple parts of the model. We evaluate our approach on multi-location model changes that have actually been performed by developers in real-world projects. NextFocus achieves promising results for multi-location model completion, even when changes are heavily spread across the model. It achieves an average Precision@k score of 0.98 for $k \leq 10$, significantly outperforming the three baseline approaches.

Figures (10)

• Figure 1: Examples of single-location and multi-location software model changes with different values of $c$ and $s$, green element mark newly added elements
• Figure 2: Combined process of single-location model completion and next focus node prediction.
• Figure 3: Overview of the global embedding-based next focus predictor (Next-Focus) approach.
• Figure 4: Neural Network architecture of Next-Focus.
• Figure 5: Precision@$k$ distribution of semantic similarity, historical co-change frequency, random selection, and Next-Focus.

Theorems & Definitions (8)

• Definition 3.1: Abstract syntax graph
• Definition 3.2: Labeled directed graph
• Definition 3.3: Direct successor set
• Definition 3.4: Structural model difference
• Definition 3.5: Model completion
• Definition 3.6: Single-location model completion
• Definition 3.7: Multi-location model completion
• Definition 4.1