What question did this study set out to answer?

The aim is to enhance the explainability of transformer models by optimizing feature corruption sequences using a novel framework called EvoDropX.

April 17, 2026Open Access

EvoDropX: Evolutionary optimization of feature corruption sequences for faithful explanations of transformer models

View Full Paper

Ask AI

Bookmark

Authors

Evaluation method demonstrates improved fidelity of explanations in transformer models, suggesting better transparency in AI systems.

The aim is to enhance the explainability of transformer models by optimizing feature corruption sequences using a novel framework called EvoDropX.
Developed EvoDropX as an optimization problem to improve explanation fidelity.
Utilized Grammatical Evolution to evolve feature corruption sequences maximizing SRG.
Conducted experiments across multiple datasets, including IMDb and Stanford Sentiment Treebank, using different transformer models.
Compared EvoDropX performance against state-of-the-art xAI methods like SHAP and LIME.
EvoDropX significantly outperforms existing xAI methods under the SRG criterion.
Achieved a 74.77% improvement in SRG on the IMDB dataset with the BERT model.
Consistent performance improvements observed across all dataset-model pairs.
Qualitative analyses indicate EvoDropX identifies sentiment-bearing terms and their relationships effectively.

Singh et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark

April 17, 2026Open Access

View Full Paper

Ask AI

Bookmark

Authors

Evaluation method demonstrates improved fidelity of explanations in transformer models, suggesting better transparency in AI systems.

The aim is to enhance the explainability of transformer models by optimizing feature corruption sequences using a novel framework called EvoDropX.
Developed EvoDropX as an optimization problem to improve explanation fidelity.
Utilized Grammatical Evolution to evolve feature corruption sequences maximizing SRG.
Conducted experiments across multiple datasets, including IMDb and Stanford Sentiment Treebank, using different transformer models.
Compared EvoDropX performance against state-of-the-art xAI methods like SHAP and LIME.
EvoDropX significantly outperforms existing xAI methods under the SRG criterion.
Achieved a 74.77% improvement in SRG on the IMDB dataset with the BERT model.
Consistent performance improvements observed across all dataset-model pairs.
Qualitative analyses indicate EvoDropX identifies sentiment-bearing terms and their relationships effectively.

Singh et al. (2026) studied this question.

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

View Full Paper

Ask AI

Bookmark