Towards Explainability in Retrieval-Augmented LLMs
Bibliographic record
Abstract
In an era where artificial intelligence (AI) is re-shaping countless aspects of society, we present a forward-looking perspective for enhancing the explainability of large language models (LLMs), with a particular focus on the retrieval-augmented generation (RAG) prompting technique. We motivate the urgency for developing techniques to explain LLM decision-making behaviour, especially as these models are deployed in critical sectors. Central to this effort is RAGE, our novel explain-ability tool that can trace the provenance of an LLM's answer back to external knowledge sources provided via RAG. RAGE builds upon established explainability techniques to recover citations for LLM answers, identify context biases, and mine answer rules. Through our novel explainability formulations and practical use cases, we chart a course toward more transparent and trustworthy AI technologies.
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 | 0.015 | 0.098 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.003 |
| Bibliometrics | 0.004 | 0.002 |
| Science and technology studies | 0.002 | 0.005 |
| Scholarly communication | 0.005 | 0.014 |
| Open science | 0.002 | 0.009 |
| Research integrity | 0.003 | 0.004 |
| Insufficient payload (model declined to judge) | 0.006 | 0.001 |
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".