OpenOneRec Technical Report: Advancing Recommender Systems
Published:Dec 31, 2025 10:15
•1 min read
•ArXiv
Analysis
This paper introduces RecIF-Bench, a new benchmark for evaluating recommender systems, along with a large dataset and open-sourced training pipeline. It also presents the OneRec-Foundation models, which achieve state-of-the-art results. The work addresses the limitations of current recommendation systems by integrating world knowledge and reasoning capabilities, moving towards more intelligent systems.
Key Takeaways
- •Proposes RecIF-Bench, a holistic benchmark for evaluating recommender systems.
- •Releases a large training dataset with 96 million interactions.
- •Open-sources a comprehensive training pipeline.
- •Introduces OneRec-Foundation models achieving SOTA results.
- •Demonstrates significant improvements on the Amazon benchmark.
Reference
“OneRec Foundation (1.7B and 8B), a family of models establishing new state-of-the-art (SOTA) results across all tasks in RecIF-Bench.”