Bridging AI and Regulation: Large Language Models for Documentation Compliance Check
Bibliographic record
Abstract
As Artificial Intelligence (AI) permeates our lives more rapidly, robust regulations to ensure trustworthy AI applications are demanded by AI operators and consumers. The European Union’s AI Act addresses this by establishing a regulatory framework for high-risk AI systems, emphasizing the need for proper documentation to ensure compliance. This paper presents a novel approach to assess AI documentation using Large Language Model based methods: a GPT-4 prompting approach and fine-tuning DeBERTa and Mistral-7B. Due to the lack of relevant datasets, we constructed a novel benchmark dataset comprising text passages from AI research publications. These passages are matched by AI experts with regulatory requirements and are classified into different fulfilment classes. Using this dataset in our comparative study, our findings demonstrate that fine-tuning the DeBERTa model achieves 92% ± 1% accuracy in classifying compliance categories, outperforming the more complex GPT-4 and Mistral-7B significantly. Overall, this research advances AI governance by providing insights into automating documentation compliance checks. Finally, by making the models and datasets publicly available, we promote further research into enhancing transparency and accountability in AI 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 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 | 0.010 | 0.051 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.004 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.005 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.004 |
| 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".