Large Language Models for API Classification: An Explorative Study
Bibliographic record
Abstract
Linking APIs to the business functions they implement is crucial for handling software operations, especially during recovery from disasters or outages. In this context, the speed and accuracy of operators in linking them impact response time during mission-critical operation activities. Besides, this linkage is essential to designing preventive actions, such as resilience strategies. Automatic API classification using Large Language Models (LLMs) may simplify and speed up APIs-business function linkage. However, previous studies unveiled the barriers practitioners face when deciding on and adopting LLMs in software engineering (SE) tasks due to a lack of guidance for non-experts. This paper aims to lower barriers to using LLMs by systems operators and site reliability engineers (SREs), focusing on the API classification task in the context of operational activities. Based on three cases from the finance industry, we extracted requirements for LLM usage, and assessed 14 recently released LLMs on this task. Our results demonstrate that LLMs accurately classify APIs using business function targets with an F1–Score of 89.5 for the leading LLM without requiring specific LLM expertise and resource-intensive fine-tuning. Besides, our findings on LLMs’ performance and reliability mark a significant advancement in comparing open and closed-source and general and domain-specific LLMs in an SE classification task. Eventually, our experiments yield practical guidance for implementing LLMs in this context. Artifacts used in and generated by the experiments are publicly available at https://bit.ly/llms4apiclassification.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.018 | 0.071 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.003 |
| Bibliometrics | 0.004 | 0.003 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.004 | 0.007 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.005 |
| Insufficient payload (model declined to judge) | 0.003 | 0.002 |
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 source (direct Gemma or distilled Codex), 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".