On the Impact of Multi-language Development in Machine Learning Frameworks
Bibliographic record
Abstract
The role of machine learning frameworks in software applications has exploded in recent years. Similar to non-machine learning frameworks, those frameworks need to evolve to incorporate new features, optimizations, etc., yet their evolution is impacted by the interdisciplinary development teams needed to develop them: scientists and developers. One concrete way in which this shows is through the use of multiple programming languages in their code base, enabling the scientists to write optimized low-level code while developers can integrate the latter into a robust framework. Since multi-language code bases have been shown to impact the development process, this paper empirically compares ten large open-source multi-language machine learning frameworks and ten large open-source multi-language traditional systems in terms of the volume of pull requests, their acceptance ratio i.e., the percentage of accepted pull requests among all the received pull requests, review process duration i.e., period taken to accept or reject a pull request, and bug-proneness. We find that multi-language pull request contributions present a challenge for both machine learning and traditional systems. Our main findings show that in both machine learning and traditional systems, multi-language pull requests are likely to be less accepted than mono-language pull requests; it also takes longer for both multi- and mono-language pull requests to be rejected than accepted. Machine learning frameworks take longer to accept/reject a multi-language pull request than traditional systems. Finally, we find that mono-language pull requests in machine learning frameworks are more bug-prone than traditional systems.
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.
How this classification was reachedexpand
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 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".