Tag: Transformer models

A hybrid TabNet and FT-Transformer model classified sports event security risk levels with 99.53% accuracy on a 40,000-record Kaggle dataset, ...

ViolenceBanglaBERT combined contextual Bangla representations, a lightweight classifier and back-translation augmentation to reach a 0.75 macro F1-score in a controlled ...

A study combining Transformer, deep neural-network and recurrent models with adaptive weighting reported 95% accuracy and 96% recall when detecting ...

A review of 2019–2025 research finds that transformer models show promise for lung cancer detection and classification, but variable validation ...

A systematic review of 87 studies identifies fragmentation across multimodal fusion, adaptive learning, edge inference, privacy and explainability as the ...

Researchers built the first extractive summarization system for Indonesian Friday sermon transcripts by fine-tuning IndoBERT on automatically generated pseudo-labels, achieving ...

A controlled benchmark of four pre-trained transformer models on the Devign dataset found CodeBERT achieved the highest mean accuracy, precision, ...

A PSO-fused model combining text, speech and video representations reached 94.0% accuracy on 690 labeled television-dialogue utterances in the MUStARD ...

CarBench evaluates 11 neural-surrogate configurations on more than 8,000 high-fidelity car simulations, comparing accuracy, pressure-field fidelity, efficiency, uncertainty and cross-category ...

A transformer-autoencoder method that groups federated clients without labels improved sentiment-classification accuracy across controlled label-skew conditions on Sentiment140 and IMDB.

A context-aware deep-learning model classified rodent sleep stages across unseen animals and retained near-baseline performance after training only on vehicle-control ...

A comparative study finds that a fine-tuned DistilBERT pipeline reliably detects deceptive hotel reviews while a Perceiver-based pipeline collapses toward ...

A new transformer model called LitGene learns gene function by reading the scientific literature and aligning it with Gene Ontology ...

A new expert-validated dataset provides multi-label artifact annotations for over 20,000 dermoscopic images, exposing the visual noise that undermines AI ...

Researchers at BRAC University developed ClusSumm, a framework combining hierarchical clustering with adaptive transformer models that compresses thousands of MOOC ...

A new hybrid framework called LLMTransRec combines transformer sequence modeling with large language model embeddings to improve recommendations across sparse ...

Researchers converted patent abstracts into semantic embeddings to measure technological similarity between firms and found the resulting network predicts subsequent ...

A new multi-domain Arabic dataset and benchmark shows that deepfake-text detectors excel within familiar domains but falter sharply on government ...

Researchers have unveiled a collaborative encryption framework that trains deep learning models on sensitive data up to three times faster ...

In the revolutionary field of proteomics, understanding the dynamics of peptide turnover has emerged as a frontline challenge for researchers. ...