Skip to content

EthenEthenEthen

LoginPlatformTry Ethen

Open Source Model Profile · XSY

albert-base-v2-fakenews-discriminator

albert-base-v2-fakenews-discriminator is an ALBERT text-classification fine-tune from XSY. According to the model card, it distinguishes fake from real news titles.

Publisher
XSY
Task
text-classification
Model type
albert
License
apache-2.0
Library
transformers
Publication status
Accepted · not indexed

Model overview

albert-base-v2-fakenews-discriminator is published by XSY as a Transformers text-classification model. The captured configuration identifies AlbertForSequenceClassification with model type albert under Apache-2.0. According to the model card, it fine-tunes albert-base-v2 on news titles labeled fake or real.

Recorded capabilities

Fake-news classifier

According to the model card, this is an albert-base-v2 fine-tune that classifies titles as fake news or real news.

Publisher-reported accuracy

According to the model card, evaluation reports 0.9758 accuracy with 0.0910 validation loss after one epoch.

ALBERT sequence-classification build

Captured configuration identifies AlbertForSequenceClassification with model type albert under Apache-2.0 licensing.

Documented training setup

The model card documents learning rate 5e-05, batch size 16, seed 42, Adam, linear scheduling with 500 warmup steps, and one epoch.

Use cases in the source record

  • Title-level fake-versus-real news classification research using the publisher's label scheme.
  • Small-classifier baseline experiments that reproduce the publisher's reported accuracy and loss setup.

Limitations and unknowns

  • Accuracy and loss figures are publisher-reported; no Ethen-measured evaluation results were extracted.
  • The model card marks intended uses, limitations, and some training and evaluation data sections as needing more information.
  • No parameter count or context-window value was extracted from this record.
  • Provider state is historical snapshot data, not independently refreshed current availability.

Source and provenance

Source: XSY/albert-base-v2-fakenews-discriminator

Captured: Unknown. Processed: 2026-09-07T19:34:38.719796+00:00.

albert-base-v2-fakenews-discriminator The dataset: Fake and real news dataset https://www.kaggle.com/clmentbisaillon/fake-and-real-news-dataset I use title and label to train the classifier label_0 : Fake news label_1 : Real news This model is a fine-tuned version of albert-base-v2 on an unknown dataset. It achieves the following results on the evaluation set: Loss: 0.0910 Accuracy: 0.9758 Model description More information needed Intended uses & limitations More information needed Training and evaluation data More information needed Training procedure Training hyperparameters The following hyperparameters were used during training:…

F001F002F003F004F005F006F009F010F011F012