Investigating Data Leakage–Induced Over-Confidence and Explanation Faithfulness in Transformer-Based Text and Audio Models
Year: 2026
Type: article
Abstract: Transformer-based models now sit at the center of modern artificial intelligence, powering systems that read, listen, respond, and increasingly act on behalf of humans. From large-scale language models to voice-driven assistants, these systems exhibit a level of fluency and responsiveness that would have seemed implausible only a few years ago. Their success has been driven largely by scale Larger models, larger datasets, and longer training regimes resulting in impressive performance across text and audio tasks.
Author: Rahul Vakiti
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W7125117459
FWCI:
Cites:
Cited by:
Related to: