A training technique where irrelevant data points are sampled to improve the performance of retrieval models.
More about Negative Sampling in Retrieval
Negative Sampling in Retrieval is a method used to train retrieval models by introducing irrelevant examples (negatives) during the learning process. This helps the model differentiate between relevant and irrelevant data points, improving its ability to rank results accurately.
Negative sampling is a critical component in training tasks like dense retrieval, semantic search, and retrieval fusion, where distinguishing relevance is essential.