Hugging Face's logo
Activity Feed

AI & ML interests

NewsReX aims to continuously incorporate emerging models and techniques, providing both the research community and industry practitioners with a reliable, up-to-date foundation for developing and benchmarking news recommendation systems.

Organization Card
Community About org cards

NewsReX Logo

A Modular, Multi-Framework Benchmark for Neural News Recommendation Research

GitHub arXiv License Python Version

Overview

NewsReX provides reproducible benchmarks for modern news recommendation algorithms. All models are implemented identically across both JAX/Flax and PyTorch, ensuring completely transparent, framework-agnostic architectural comparisons under identical training conditions.

Implemented Models

Model Venue JAX / Flax PyTorch Paper Link
NRMS EMNLP 2019 ✅ ✅ Wu et al.
NAML IJCAI 2019 ✅ ✅ Wu et al.
LSTUR ACL 2019 ✅ ✅ An et al.
MINER ACL Findings 2022 ✅ ✅ Li et al.
PP-Rec ACL 2021 ✅ ✅ Qi et al.
CAUM SIGIR 2022 ✅ ✅ Qi et al.
CROWN WWW 2025 ✅ ✅ Ko et al.
DIGAT EMNLP Findings 2022 ✅ ✅ Mao et al.
GLORY RecSys 2023 ✅ ✅ Yang et al.
TCCM CIKM 2023 ✅ ✅ Qi et al.

⚡ Quick Start

# Clone the architecture framework
git clone [https://github.com/igor17400/NewsReX.git](https://github.com/igor17400/NewsReX.git)
cd NewsReX

# Manage environment and sync dependencies smoothly with uv
uv sync --extra all-cuda

# Train a chosen target baseline (e.g., NRMS using JAX backend)
uv run python src/train.py experiment=mind/nrms framework=jax

# Stream pre-trained model weights straight out of the Hub zoo
uv run python src/train.py experiment=mind/nrms framework=jax \
    weights=hf://newsrex/NRMS-JAX-MIND-small/model.safetensors

datasets 0

None public yet