Younggue Bae

In-Batch Negatives Can Silently Cripple LLM-Encoded Sequential Recommenders

LLM-based sequential recommenders increasingly train user/item encoders with in-batch InfoNCE negatives inherited from contrastive learning—convenient, but it silently caps the negative pool at batch size rather than catalog size whenever encoder cost bounds the batch. We report an early finding: a 15-negative in-batch...