What question did this study set out to answer?

To develop and evaluate a hybrid explainable AI framework combining saliency attribution, causal reasoning, and user visualization to improve transparency in large language models.

September 17, 2026PLoS ONEOpen Access

Bridging trust and performance in intelligent systems: Hybrid explainable AI approaches for interpreting large language models

View Full Paper

Ask AI

Bookmark

Authors

APArul Selvam P.Hindustan Institute of Technology and Science TPTamije Selvy P.Hindustan Institute of Technology and Science

Discussion

Member takes

Overview

Benchmarking study demonstrates enhanced explanation fidelity in large language models, indicating viable pathways for transparent AI adoption.

Key Points

To develop and evaluate a hybrid explainable AI framework combining saliency attribution, causal reasoning, and user visualization to improve transparency in large language models.
Evaluated the unified pipeline across multiple architectures (BERT, T5, GPT, and LLaMA) on benchmark datasets, including GLUE, SQuAD, IMDB, and domain-specific corpora.
Quantified explanation fidelity using standardized insertion and deletion metrics alongside human-centered user evaluations for clarity and trust.
Conducted domain-specific case studies examining sentiment analysis and question answering.
Demonstrated up to a 15-20% improvement in explanation fidelity compared to standalone attention-based approaches based on insertion and deletion metrics.
Achieved higher user clarity and trust ratings while adding less than 25% computational overhead.
Outperformed existing baselines in generating intuitive and precise reasoning paths in sentiment analysis and question answering tasks.

Cite This Study

P. et al. (2026) studied this question.

synapsesocial.com/papers/6aabb73d5f706d05830e627f https://doi.org/10.1371/journal.pone.0343472

View Full Paper

Ask AI

Bookmark

September 17, 2026PLoS ONEOpen Access

Bridging trust and performance in intelligent systems: Hybrid explainable AI approaches for interpreting large language models

View Full Paper

Ask AI

Bookmark

Authors

APArul Selvam P.Hindustan Institute of Technology and Science TPTamije Selvy P.Hindustan Institute of Technology and Science

Discussion

Member takes

Overview

Benchmarking study demonstrates enhanced explanation fidelity in large language models, indicating viable pathways for transparent AI adoption.

Key Points

To develop and evaluate a hybrid explainable AI framework combining saliency attribution, causal reasoning, and user visualization to improve transparency in large language models.
Evaluated the unified pipeline across multiple architectures (BERT, T5, GPT, and LLaMA) on benchmark datasets, including GLUE, SQuAD, IMDB, and domain-specific corpora.
Quantified explanation fidelity using standardized insertion and deletion metrics alongside human-centered user evaluations for clarity and trust.
Conducted domain-specific case studies examining sentiment analysis and question answering.
Demonstrated up to a 15-20% improvement in explanation fidelity compared to standalone attention-based approaches based on insertion and deletion metrics.
Achieved higher user clarity and trust ratings while adding less than 25% computational overhead.
Outperformed existing baselines in generating intuitive and precise reasoning paths in sentiment analysis and question answering tasks.

Cite This Study

P. et al. (2026) studied this question.

synapsesocial.com/papers/6aabb73d5f706d05830e627f https://doi.org/10.1371/journal.pone.0343472

Also Consider

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

View Full Paper

Ask AI

Bookmark