BLAZE: Cross-Language and Cross-Project Bug Localization via Dynamic Chunking and Hard Example Learning
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Software bugs require developers to expend significant effort to identify and resolve them, often consuming about one-third of their time. Bug localization, the process of pinpointing the exact source code files that need modification, is crucial in reducing this effort. Existing bug localization tools, typically reliant on deep learning techniques, face limitations in both cross-project applicability and multi-language environments. <p xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Recent advancements with Large Language Models (LLMs) offer detailed representations for bug localization that may help to overcome such limitations. However, these models are known to encounter challenges with 1) limited context windows and 2) mapping accuracy. To address these challenges, we propose <monospace>BLAZE</monospace>, an approach that employs <i>dynamic chunking</i> and <i>hard example learning</i>. First, <monospace>BLAZE</monospace> dynamically segments source code to minimize continuity loss. Then, <monospace>BLAZE</monospace> fine-tunes a GPT-based model using complex bug reports in order to enhance cross-project and cross-language bug localization. To support the capability of <monospace>BLAZE</monospace>, we create the <monospace>BeetleBox</monospace> dataset, which comprises 23,782 bugs from 29 large and thriving opensource projects across five programming languages (Java, C++, Python, Go, and JavaScript). Our evaluation of <monospace>BLAZE</monospace> on three benchmark datasets—<monospace>BeetleBox</monospace>, SWE-Bench, and Ye et al.—demonstrates substantial improvements compared to six <i>state-of-the-art</i> baselines. Specifically, <monospace>BLAZE</monospace> achieves up to an increase of 120% in Top 1 accuracy, 144% in Mean Average Precision (MAP), and 100% in Mean Reciprocal Rank (MRR). Furthermore, an extensive ablation study confirms the contributions of our pipeline components to the overall performance enhancement.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it