A real-time hard negative sampling technique using LLM-based clustering outperforms standard in-batch and out-of-batch methods for training two-tower models in large-scale recommendation systems.
arXiv preprint arXiv:2010.03240 , year=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Real-Time Hard Negative Sampling via LLM-based Clustering for Large-Scale Two-Tower Retrieval
A real-time hard negative sampling technique using LLM-based clustering outperforms standard in-batch and out-of-batch methods for training two-tower models in large-scale recommendation systems.