EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Evaluating explainable AI attribution methods in neural machine translation via attention-guided knowledge distillation

Aria Nourbakhsh, Salima Lamsiyah, Adelaide Danilov, Christoph Schommer

Natural Language Processing · 2026-05 · 32 (3) · 267-301

DOI: 10.1017/nlp.2026.10028 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Natural Language Processing, online erschienen am 28.05.2026. Verfasst von Aria Nourbakhsh und 3 weiteren Autor:innen. Die Arbeit nennt 81 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The study of the attribution of input features to the output of neural network models is an active area of research. While numerous Explainable AI (XAI) techniques have been proposed to interpret these models, the systematic and automated evaluation of these methods in sequence-to-sequence (seq2seq) models is less explored. This paper introduces a new approach for evaluating explainability methods in transformer-based seq2seq models, building upon the forward simulation of XAI methods. We use teacher-derived attribution maps as a structured side signal to guide a student model, and quantify the utility of different attribution methods through the student’s ability to simulate targets. Using the Inseq library, we extract attribution scores over source–target sequence pairs and inject these scores into the attention mechanism of a student transformer model under four composition operators (addition, multiplication, averaging, and replacement). Across three language pairs (de–en, fr–en, ar–en) and attributions from Marian-MT and mBART models, Attention, Value Zeroing, and Layer Gradient times × $\times$ Activation consistently yield the largest gains in BLEU (and corresponding improvements in chrF) relative to baselines. In contrast, other gradient-based methods (Saliency, Integrated Gradients, DeepLIFT, Input times × $\times$ Gradient, GradientShap) lead to smaller and less consistent improvements. These results suggest that different attribution methods capture distinct signals and that attention-derived attributions better capture the alignment between source and target representations in seq2seq models. Finally, we introduce an Attributor transformer that, given a source–target pair, learns to reconstruct the teacher’s attribution map. Our findings demonstrate that the more accurately the Attributor can reproduce attribution maps, the more useful an injection of those maps is for the downstream task. The source code can be found on GitHub. a

Zitieren

APA

Nourbakhsh, A., Lamsiyah, S., Danilov, A., & Schommer, C. (2026). Evaluating explainable AI attribution methods in neural machine translation via attention-guided knowledge distillation. Natural Language Processing, 32(3), 267-301. https://doi.org/10.1017/nlp.2026.10028

Vancouver

Nourbakhsh A, Lamsiyah S, Danilov A, Schommer C. Evaluating explainable AI attribution methods in neural machine translation via attention-guided knowledge distillation. Natural Language Processing. 2026;32(3):267-301. doi:10.1017/nlp.2026.10028

BibTeX
@article{Nourbakhsh2026,
  author = {Nourbakhsh, Aria and Lamsiyah, Salima and Danilov, Adelaide and Schommer, Christoph},
  title = {Evaluating explainable AI attribution methods in neural machine translation via attention-guided knowledge distillation},
  journal = {Natural Language Processing},
  year = {2026},
  volume = {32},
  number = {3},
  pages = {267--301},
  doi = {10.1017/nlp.2026.10028}
}
Kostenlos starten →