Real-Time Hard Negative Sampling via LLM-based Clustering for Large-Scale Two-Tower Retrieval
For large-scale recommendation systems, this method addresses the problem of easy negatives in training, offering a practical solution to improve retrieval quality and mitigate feedback loops.
The paper proposes a self-supervised hard negative sampling technique using LLM-based clustering for two-tower retrieval models, which outperforms industry-standard methods in experiments and online deployment, reducing popularity bias.
The two-tower model has been widely used for large-scale recommendation systems, particularly in the retrieval stage. Industry standards for training two-tower models typically involve in-batch and/or out-of-batch negative sampling. However, these methods often produce easy negatives that models can quickly learn, failing to sufficiently challenge the model. To address this issue, a novel self-supervised hard negative sampling technique is proposed that leverages a large language model (LLM) to generate hard negatives from the same cluster during model training. By utilizing the LLM to learn media representations, the proposed approach ensures that the generated negatives are more challenging and informative. This real-time sampling framework is designed for seamless integration into production models, capable of handling billions of training data points with minimal computational complexity. Experiments on public datasets, along with deployment to a large-scale online system, demonstrate that the proposed negative sampling technique outperforms widely used industry methods. Furthermore, analysis in industrial applications reveals that this sampling method can help break inherent feedback loops in recommendations and significantly reduce popularity bias.