Ko-Jen Hsiao

Keertana Chidambaram, Sanath Kumar Krishnamurthy, Qiuling Xu et al.

Aligning generative recommender systems to user preferences via post-training is critical for closing the gap between next-item prediction and actual recommendation quality. Existing post-training methods are ill-suited for production-scale systems: RLHF methods reward hack due to noisy user feedback and unreliable reward models, offline RL alternatives require propensity scores that are unavailable, and online interaction is infeasible. We identify exponential reward-weighted SFT with weights $w = \exp(r/λ)$ as uniquely suited to this setting, and provide the theoretical and empirical foundations that explain why. By optimizing directly on observed rewards without querying a learned reward model, the method is immune to reward hacking, requires no propensity scores, and is fully offline. We prove the first policy improvement guarantees for this setting under noisy rewards, showing that the gap scales only logarithmically with catalog size and remains informative even for large item catalogs. Crucially, we show that temperature $λ$ explicitly and quantifiably controls the robustness-improvement tradeoff, providing practitioners with a single interpretable regularization hyperparameter with theoretical grounding. Experiments on three open-source and one proprietary dataset against four baselines confirm that exponential reward weighting is simple, scalable, and consistently outperforms RLHF-based alternatives.

Qiuling Xu, Ko-Jen Hsiao, Moumita Bhattacharya

Generative recommendation models can model user behavior as sequences of events and provide a shared backbone for multiple recommendation tasks. In production, however, pre-training gains do not automatically translate into downstream application improvements: task headroom, repeated-training cost, serving latency, and item freshness all affect transfer. We describe our experience scaling a generative recommender from 2M to 1B backbone parameters, excluding embedding and decoding layers, in a production-scale title recommendation setting. Across multiple downstream tasks, we observe task-dependent scaling behavior: some tasks approach an empirical ceiling within the observed scale range, while others continue to benefit from additional capacity. This motivates using offset scaling-law fits as a diagnostic for where additional model scale may be more or less useful. We then study production constraints that arise when applying the model in practice. Frequent retraining over trillions of behavior tokens makes training and decoding efficiency important; cached serving can make the immediate next-token target stale; and newly launched titles may need to be scored from semantic metadata before collaborative ID embeddings are reliable. We address these issues with multi-token prediction for serving-latency alignment, sampled softmax and a projected decoding head for efficient repeated training, and semantic item towers with collaborative-embedding masking for cold-start adaptation. In a one-week production-shadow evaluation over 1M users, the 1B-backbone model achieves higher MRR than the 2M-backbone baseline across all reported tasks. Overall, the results support treating model scale as one component of a production transfer problem, alongside task headroom, decoding cost, serving-latency alignment, and item generalization.

Swanand Joshi, Yesu Feng, Ko-Jen Hsiao et al.

Long-lived recommender systems (RecSys) often encounter lengthy user-item interaction histories that span many years. To effectively learn long term user preferences, Large RecSys foundation models (FM) need to encode this information in pretraining. Usually, this is done by either generating a long enough sequence length to take all history sequences as input at the cost of large model input dimension or by dropping some parts of the user history to accommodate model size and latency requirements on the production serving side. In this paper, we introduce a sliding window training technique to incorporate long user history sequences during training time without increasing the model input dimension. We show the quantitative & qualitative improvements this technique brings to the RecSys FM in learning user long term preferences. We additionally show that the average quality of items in the catalog learnt in pretraining also improves.

Ko-Jen Hsiao, Kevin S. Xu, Jeff Calder et al.

We consider the problem of identifying patterns in a data set that exhibit anomalous behavior, often referred to as anomaly detection. Similarity-based anomaly detection algorithms detect abnormally large amounts of similarity or dissimilarity, e.g.~as measured by nearest neighbor Euclidean distances between a test sample and the training samples. In many application domains there may not exist a single dissimilarity measure that captures all possible anomalous patterns. In such cases, multiple dissimilarity measures can be defined, including non-metric measures, and one can test for anomalies by scalarizing using a non-negative linear combination of them. If the relative importance of the different dissimilarity measures are not known in advance, as in many anomaly detection applications, the anomaly detection algorithm may need to be executed multiple times with different choices of weights in the linear combination. In this paper, we propose a method for similarity-based anomaly detection using a novel multi-criteria dissimilarity measure, the Pareto depth. The proposed Pareto depth analysis (PDA) anomaly detection algorithm uses the concept of Pareto optimality to detect anomalies under multiple criteria without having to run an algorithm multiple times with different choices of weights. The proposed PDA approach is provably better than using linear combinations of the criteria and shows superior performance on experiments with synthetic and real data sets.

Ko-Jen Hsiao, Alex Kulesza, Alfred Hero

Socially-based recommendation systems have recently attracted significant interest, and a number of studies have shown that social information can dramatically improve a system's predictions of user interests. Meanwhile, there are now many potential applications that involve aspects of both recommendation and information retrieval, and the task of collaborative retrieval---a combination of these two traditional problems---has recently been introduced. Successful collaborative retrieval requires overcoming severe data sparsity, making additional sources of information, such as social graphs, particularly valuable. In this paper we propose a new model for collaborative retrieval, and show that our algorithm outperforms current state-of-the-art approaches by incorporating information from social networks. We also provide empirical analyses of the ways in which cultural interests propagate along a social graph using a real-world music dataset.

Ko-Jen Hsiao, Jeff Calder, Alfred O. Hero

Most content-based image retrieval systems consider either one single query, or multiple queries that include the same object or represent the same semantic information. In this paper we consider the content-based image retrieval problem for multiple query images corresponding to different image semantics. We propose a novel multiple-query information retrieval algorithm that combines the Pareto front method (PFM) with efficient manifold ranking (EMR). We show that our proposed algorithm outperforms state of the art multiple-query retrieval algorithms on real-world image databases. We attribute this performance improvement to concavity properties of the Pareto fronts, and prove a theoretical result that characterizes the asymptotic concavity of the fronts.