Search papers, labs, and topics across Lattice.
This paper introduces a knowledge distillation-based approach to evaluate XAI attribution methods in transformer-based seq2seq models. Attribution maps from a teacher model are injected into the attention mechanism of a student model, and the student's BLEU score is used to quantify the utility of different attribution methods. Experiments across three language pairs show that Attention, Value Zeroing, and Layer Gradient x Activation consistently yield the largest gains in BLEU, suggesting they better capture source-target alignment.
Forget subjective human evaluations: this paper uses a clever knowledge distillation trick to objectively rank XAI methods for NMT, revealing that attention-based attributions beat gradient-based ones.
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. 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 $\times$ 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 $\times$ 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 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.
