While large transformer models have been successfully used in many real-world applications such as natural language processing, computer vision, and speech processing, scaling transformers for recommender systems remains a challenging problem. Recently, the Generative Recommenders framework was proposed as a way to scale beyond typical Deep Learning Recommendation Models (DLRMs). By reformulating recommendation as a sequential transduction task, it improves scaling properties in terms of compute. Nevertheless, the largest encoder configuration reported by the HSTU authors is only 176 million parameters --- far smaller than the hundreds of billions (or even trillions) that are now common in language models.

Generative retrieval with Semantic IDs (SIDs) assigns each item a discrete identifier and treats retrieval as a sequence generation problem rather than a nearest-neighbor search. While content-only SIDs are stable, they do not take into account user-item interaction patterns, so recent systems construct interaction-informed SIDs. However, as interaction patterns drift over time, these identifiers become stale, i.e., their collaborative semantics no longer match recent logs.

Two-tower neural networks are a popular architecture for the retrieval stage in recommender systems. These models are typically trained with a softmax loss over the item catalog. However, in web-scale settings, the item catalog is often prohibitively large, making full softmax infeasible. A common solution is sampled softmax, which approximates the full softmax using a small number of sampled negatives. One practical and widely adopted approach is to use in-batch negatives, where negatives are drawn from items in the current mini-batch. However, this introduces a bias: items that appear more frequently in the batch (i.e., popular items) are penalized more heavily. To mitigate this issue, a popular industry technique known as logQ correction adjusts the logits during training by subtracting the log-probability of an item appearing in the batch. This correction is derived by analyzing the bias in the gradient and applying importance sampling, effectively twice, using the in-batch distribution as a proposal distribution. While this approach improves model quality, it does not fully eliminate the bias. In this work, we revisit the derivation of logQ correction and show that it overlooks a subtle but important detail: the positive item in the denominator is not Monte Carlo-sampled - it is always present with probability 1. We propose a refined correction formula that accounts for this. Notably, our loss introduces an interpretable sample weight that reflects the model's uncertainty - the probability of misclassification under the current parameters. We evaluate our method on both public and proprietary datasets, demonstrating consistent improvements over the standard logQ correction.