jibala-1022/easyrec-large

EasyRec-Base

Overview

• Description: EasyRec is a series of language models designed for recommendations, trained to match the textual profiles of users and items with collaborative signals.
• Usage: You can use EasyRec to encode user and item text embeddings based on the textual profiles that reflect their preferences for various recommendation scenarios.
• Evaluation: We evaluate the performance of EasyRec in: (i) Text-based zero-shot recommendation and (ii) Text-enhanced collaborative filtering.
• Finetuned from model: EasyRec is finetuned from RoBERTa within English.

For details please refer [💻GitHub Code] and [📖Paper].

Model List

We release a series of EasyRec checkpoints with varying sizes. You can easily load these models from Hugging Face by replacing the model name.

Model	Size	Parameters	Recall@10 on Movies
jibala-1022/easyrec-small	243 MB	121,364,313	0.0086
jibala-1022/easyrec-base	328 MB	163,891,545	0.0166
jibala-1022/easyrec-large	816 MB	407,933,017	0.0166

🌟 Citation

@article{ren2024easyrec,
  title={EasyRec: Simple yet Effective Language Models for Recommendation},
  author={Ren, Xubin and Huang, Chao},
  journal={arXiv preprint arXiv:2408.08821},
  year={2024}
}