Chenglei Niu

Alexandre Salle, Chenglei Niu, Suchismit Mahapatra et al.

Personalized discovery systems often train separate models for item ranking, carousel ranking, and search, even though these tasks expose complementary signals from the same viewer journey: watches shape carousel and item ranking, search queries reveal intent even when they do not lead to a catalog match, and watch history helps interpret search as rewatching, continuation, or new discovery. We introduce the user story, a serialized representation that turns a user's cross-surface history - attributes, sessions, watch events with surface and carousel context, and search events - into a single token sequence. By interleaving pretrained language tokens with domain-specific event tokens, user stories let heterogeneous recommendation and search tasks be expressed as prompted next-token prediction over a shared grammar. TubiFM is one instantiation of this approach: a Llama 3.2 1B-based model trained on user stories and prompted to rank items, carousels, or search results without task-specific architectures. In offline evaluation, this single model outperforms specialist baselines across item, carousel, and search ranking. In online A/B tests, TubiFM significantly improves search total viewing time (TVT) by $+3.9\%$ and carousel TVT by $+0.30\%$. Item ranking is statistically neutral on TVT ($+0.14\%$), but matches a mature production stack; across all three tasks, TubiFM serves on L40S GPUs and reduces p99 ranking latency from 500ms to 200ms. These results show that shared user stories can improve discovery while simplifying ranking systems.

Yan Zheng, Qiang Chen, Chenglei Niu

Modern recommendation systems aim to increase click-through rates (CTR) for better user experience, through commonly treating ranking as a classification task focused on predicting CTR. However, there is a gap between this method and the actual objectives of businesses across different sectors. In video recommendation services, the objective of video on demand (VOD) extends beyond merely encouraging clicks, but also guiding users to discover their true interests, leading to increased watch time. And longer users watch time will leads to more revenue through increased chances of presenting online display advertisements. This research addresses the issue by redefining the problem from classification to regression, with a focus on maximizing revenue through user viewing time. Due to the lack of positive labels on recommendation, the study introduces Tweedie Loss Function, which is better suited in this scenario than the traditional mean square error loss. The paper also provides insights on how Tweedie process capture users diverse interests. Our offline simulation and online A/B test revealed that we can substantially enhance our core business objectives: user engagement in terms of viewing time and, consequently, revenue. Additionally, we provide a theoretical comparison between the Tweedie Loss and the commonly employed viewing time weighted Logloss, highlighting why Tweedie Regression stands out as an efficient solution. We further outline a framework for designing a loss function that focuses on a singular objective.

Chenglei Niu, Guojing Zhong, Ying Liu et al.

With the rapid development of online advertising and recommendation systems, click-through rate prediction is expected to play an increasingly important role.Recently many DNN-based models which follow a similar Embedding&MLP paradigm have been proposed, and have achieved good result in image/voice and nlp fields. In these methods the Wide&Deep model announced by Google plays a key role.Most models first map large scale sparse input features into low-dimensional vectors which are transformed to fixed-length vectors, then concatenated together before being fed into a multilayer perceptron (MLP) to learn non-linear relations among input features. The number of trainable variables normally grow dramatically the number of feature fields and the embedding dimension grow. It is a big challenge to get state-of-the-art result through training deep neural network and embedding together, which falls into local optimal or overfitting easily. In this paper, we propose an Structured Semantic Model (SSM) to tackles this challenge by designing a orthogonal base convolution and pooling model which adaptively learn the multi-scale base semantic representation between features supervised by the click label.The output of SSM are then used in the Wide&Deep for CTR prediction.Experiments on two public datasets as well as real Weibo production dataset with over 1 billion samples have demonstrated the effectiveness of our proposed approach with superior performance comparing to state-of-the-art methods.