Sequential recommendation aims to predict the next item based on historical interactions. To further enhance the reasoning capability in sequential recommendation, LLMs are employed to predict the next item or generate semantic IDs for item representation, given LLMs' extensive domain knowledge and reasoning ability. However, existing LLM-based methods suffer from two limitations. (i) The scarcity of recommendation data with reasoning paths makes it challenging to design suitable chain-of-thought prompting templates, and the full potential of LLMs' reasoning abilities remains underutilized. (ii) Upon obtaining semantic IDs, the LLMs and their representations are excluded from the subsequent recommendation model training, preventing downstream models from fully utilizing the rich semantic information encoded within these IDs. To address these issues, we propose a novel CoderRec framework, which is capable of fully exploiting the information encoded in semantic IDs to guide the recommendation process. Specifically, to address the problem of scarcity in reasoning path-augmented data, we introduce latent reasoning into sequential recommendation and treat the representation captured by the downstream model as domain-specific latent thought, enabling implicit logical inference without requiring explicit CoT annotations. To ensure that the downstream recommendation models are able to deeply leverage the semantic information within IDs, we propose a novel cross-scale model collaboration strategy, which employs cross-scale IDs and a two-phase approach to align LLM-derived semantics with recommendation objectives. Extensive experiments have shown the effectiveness of our proposed CoderRec framework.

Sequential recommendation models often struggle to capture latent periodic patterns in user interests, primarily due to the noise inherent in time-domain behavioral data. While frequency-domain analysis offers a global perspective to address this, existing approaches typically treat user sequences in isolation, overlooking the crucial context of the target item. In this work, we present a novel empirical observation: user attention scores exhibit distinct spectral entropy distributions when conditioned on positive versus negative target items. Specifically, true user interests manifest as highly concentrated spectral patterns with lower entropy in the frequency domain, whereas irrelevant behaviors appear as high-entropy noise. Leveraging this insight, we propose the Frequency-Enhanced Deep Interest Network (FEDIN). FEDIN introduces a frequency-domain branch that utilizes a target-aware spectrum filtering mechanism to isolate these periodic interest signals. Extensive experiments on three public datasets demonstrate that FEDIN consistently outperforms state-of-the-art sequential recommendation baselines, demonstrating superior robustness against noise. We have released our code at: https://github.com/otokoneko/FEDIN.

Sequential recommendation (SR) aims to capture users' dynamic interests and sequential patterns based on their historical interactions. Recently, the powerful capabilities of large language models (LLMs) have driven their adoption in SR. However, we identify two critical challenges in existing LLM-based SR methods: 1) embedding collapse when incorporating pre-trained collaborative embeddings and 2) catastrophic forgetting of quantized embeddings when utilizing semantic IDs. These issues dampen the model scalability and lead to suboptimal recommendation performance. Therefore, based on LLMs like Llama3-8B-instruct, we introduce a novel SR framework named MME-SID, which integrates multimodal embeddings and quantized embeddings to mitigate embedding collapse. Additionally, we propose a Multimodal Residual Quantized Variational Autoencoder (MM-RQ-VAE) with maximum mean discrepancy as the reconstruction loss and contrastive learning for alignment, which effectively preserve intra-modal distance information and capture inter-modal correlations, respectively. To further alleviate catastrophic forgetting, we initialize the model with the trained multimodal code embeddings. Finally, we fine-tune the LLM efficiently using LoRA in a multimodal frequency-aware fusion manner. Extensive experiments on three public datasets validate the superior performance of MME-SID thanks to its capability to mitigate embedding collapse and catastrophic forgetting. The implementation code and datasets are publicly available for reproduction: https://github.com/Applied-Machine-Learning-Lab/MME-SID.

Multi-domain learning (MDL) has become a prominent topic in enhancing the quality of personalized services. It's critical to learn commonalities between domains and preserve the distinct characteristics of each domain. However, this leads to a challenging dilemma in MDL. On the one hand, a model needs to leverage domain-aware modules such as experts or embeddings to preserve each domain's distinctiveness. On the other hand, real-world datasets often exhibit long-tailed distributions across domains, where some domains may lack sufficient samples to effectively train their specific modules. Unfortunately, nearly all existing work falls short of resolving this dilemma. To this end, we propose a novel Cross-experts Covariance Loss for Disentangled Learning model (Crocodile), which employs multiple embedding tables to make the model domain-aware at the embeddings which consist most parameters in the model, and a covariance loss upon these embeddings to disentangle them, enabling the model to capture diverse user interests among domains. Empirical analysis demonstrates that our method successfully addresses both challenges and outperforms all state-of-the-art methods on public datasets. During online A/B testing in Tencent's advertising platform, Crocodile achieves 0.72% CTR lift and 0.73% GMV lift on a primary advertising scenario. The code is openly accessible at: https://github.com/SkylerLinn/Crocodile.

Large Language Models for Recommendation (LLM4Rec) have shown great potential. Many LLM4Rec approaches technically leverage a learnable tokenizer to assign item identifiers and then enable a Recommender LLM (RecLLM) to process tokenized items and user interactions for recommendation. However, a key challenge in their real-world deployment is the need for continuous retraining over time to accommodate new items and evolving user interests. While existing retraining methods can be applied to RecLLMs, learnable tokenizers introduce additional retraining challenges. We conduct a comprehensive investigation into the joint retraining of RecLLMs and learnable tokenizers, identifying key issues such as identifier collision and identifier shifts across periods. To address these, we propose Reformer, an incremental learning framework to fine-tune RecLLMs and learnable tokenizers at each period. Reformer employs a dynamic codebook to mitigate identifier collision by appending new codes and enforcing a diversity-oriented code assignment constraint. Additionally, Reformer adopts an identifier freezing strategy to ensure the invariance of previously assigned item identifiers across retraining periods. We instantiate Reformer on two representative RecLLMs and conduct extensive experiments on three real-world datasets. Substantial results demonstrate its superior retraining performance, facilitating the real-world deployment of LLM4Rec.

Lifelong user behavior sequences are crucial for capturing user interests and predicting user responses in modern recommendation systems. A two-stage paradigm is typically adopted to handle these long sequences: a subset of relevant behaviors is first searched from the original long sequences via an attention mechanism in the first stage and then aggregated with the target item to construct a discriminative representation for prediction in the second stage. In this work, we identify and characterize, for the first time, a neglected deficiency in existing long-sequence recommendation models: a single set of embeddings struggles with learning both attention and representation, leading to interference between these two processes. Initial attempts to address this issue with some common methods (e.g., linear projections---a technique borrowed from language processing) proved ineffective, shedding light on the unique challenges of recommendation models. To overcome this, we propose the Decoupled Attention and Representation Embeddings (DARE) model, where two distinct embedding tables are initialized and learned separately to fully decouple attention and representation. Extensive experiments and analysis demonstrate that DARE provides more accurate searches of correlated behaviors and outperforms baselines with AUC gains up to 9‰ on public datasets and notable improvements on Tencent's advertising platform. Furthermore, decoupling embedding spaces allows us to reduce the attention embedding dimension and accelerate the search procedure by 50\% without significant performance impact, enabling more efficient, high-performance online serving. Code in PyTorch for experiments, including model analysis, is available at https://github.com/thuml/DARE.

Click-Through Rate (CTR) prediction, a core task in recommendation systems, aims to estimate the probability of users clicking on items. Existing models predominantly follow a discriminative paradigm, which relies heavily on explicit interactions between raw ID embeddings. However, this paradigm inherently renders them susceptible to two critical issues: embedding dimensional collapse and information redundancy, stemming from the over-reliance on feature interactions over raw ID embeddings. To address these limitations, we propose a novel Supervised Feature Generation (SFG) framework, shifting the paradigm from discriminative "feature interaction" to generative "feature generation". Specifically, SFG comprises two key components: an Encoder that constructs hidden embeddings for each feature, and a Decoder tasked with regenerating the feature embeddings of all features from these hidden representations. Unlike existing generative approaches that adopt self-supervised losses, we introduce a supervised loss to utilize the supervised signal, i.e., click or not, in the CTR prediction task. This framework exhibits strong generalizability: it can be seamlessly integrated with most existing CTR models, reformulating them under the generative paradigm. Extensive experiments demonstrate that SFG consistently mitigates embedding collapse and reduces information redundancy, while yielding substantial performance gains across various datasets and base models. The code is available at https://github.com/USTC-StarTeam/GE4Rec.

Sequential Recommendation (SR) aims to leverage the sequential patterns in users' historical interactions to accurately track their preferences. However, the primary reliance of existing SR methods on collaborative data results in challenges such as the cold-start problem and sub-optimal performance. Concurrently, despite the proven effectiveness of large language models (LLMs), their integration into commercial recommender systems is impeded by issues such as high inference latency, incomplete capture of all distribution statistics, and catastrophic forgetting. To address these issues, we introduce a novel Pre-train, Align, and Disentangle (PAD) framework to enhance SR models with LLMs. In particular, we initially pre-train both the SR and LLM models to obtain collaborative and textual embeddings. Subsequently, we propose a characteristic recommendation-anchored alignment loss using multi-kernel maximum mean discrepancy with Gaussian kernels. Lastly, a triple-experts architecture, comprising aligned and modality-specific experts with disentangled embeddings, is fine-tuned in a frequency-aware manner. Experimental results on three public datasets validate the efficacy of PAD, indicating substantial enhancements and compatibility with various SR backbone models, particularly for cold items. The code and datasets are accessible for reproduction: https://github.com/Applied-Machine-Learning-Lab/PAD.

Multi-task learning (MTL) has gained significant popularity in recommender systems as it enables simultaneous optimization of multiple objectives. A key challenge in MTL is negative transfer, but existing studies explored negative transfer on all samples, overlooking the inherent complexities within them. We split the samples according to the relative amount of positive feedback among tasks. Surprisingly, negative transfer still occurs in existing MTL methods on samples that receive comparable feedback across tasks. Existing work commonly employs a shared-embedding paradigm, limiting the ability of modeling diverse user preferences on different tasks. In this paper, we introduce a novel Shared and Task-specific EMbeddings (STEM) paradigm that aims to incorporate both shared and task-specific embeddings to effectively capture task-specific user preferences. Under this paradigm, we propose a simple model STEM-Net, which is equipped with an All Forward Task-specific Backward gating network to facilitate the learning of task-specific embeddings and direct knowledge transfer across tasks. Remarkably, STEM-Net demonstrates exceptional performance on comparable samples, achieving positive transfer. Comprehensive evaluation on three public MTL recommendation datasets demonstrates that STEM-Net outperforms state-of-the-art models by a substantial margin. Our code is released at https://github.com/LiangcaiSu/STEM.

Multi-domain learning (MDL) aims to train a model with minimal average risk across multiple overlapping but non-identical domains. To tackle the challenges of dataset bias and domain domination, numerous MDL approaches have been proposed from the perspectives of seeking commonalities by aligning distributions to reduce domain gap or reserving differences by implementing domain-specific towers, gates, and even experts. MDL models are becoming more and more complex with sophisticated network architectures or loss functions, introducing extra parameters and enlarging computation costs. In this paper, we propose a frustratingly easy and hyperparameter-free multi-domain learning method named Decoupled Training (D-Train). D-Train is a tri-phase general-to-specific training strategy that first pre-trains on all domains to warm up a root model, then post-trains on each domain by splitting into multi-heads, and finally fine-tunes the heads by fixing the backbone, enabling decouple training to achieve domain independence. Despite its extraordinary simplicity and efficiency, D-Train performs remarkably well in extensive evaluations of various datasets from standard benchmarks to applications of satellite imagery and recommender systems.

Recent advances in foundation models have led to a promising trend of developing large recommendation models to leverage vast amounts of available data. Still, mainstream models remain embarrassingly small in size and naive enlarging does not lead to sufficient performance gain, suggesting a deficiency in the model scalability. In this paper, we identify the embedding collapse phenomenon as the inhibition of scalability, wherein the embedding matrix tends to occupy a low-dimensional subspace. Through empirical and theoretical analysis, we demonstrate a two-sided effect of feature interaction specific to recommendation models. On the one hand, interacting with collapsed embeddings restricts embedding learning and exacerbates the collapse issue. On the other hand, interaction is crucial in mitigating the fitting of spurious features as a scalability guarantee. Based on our analysis, we propose a simple yet effective multi-embedding design incorporating embedding-set-specific interaction modules to learn embedding sets with large diversity and thus reduce collapse. Extensive experiments demonstrate that this proposed design provides consistent scalability and effective collapse mitigation for various recommendation models. Code is available at this repository: https://github.com/thuml/Multi-Embedding.

Click-through rate (CTR) prediction is a crucial area of research in online advertising. While binary cross entropy (BCE) has been widely used as the optimization objective for treating CTR prediction as a binary classification problem, recent advancements have shown that combining BCE loss with an auxiliary ranking loss can significantly improve performance. However, the full effectiveness of this combination loss is not yet fully understood. In this paper, we uncover a new challenge associated with the BCE loss in scenarios where positive feedback is sparse: the issue of gradient vanishing for negative samples. We introduce a novel perspective on the effectiveness of the auxiliary ranking loss in CTR prediction: it generates larger gradients on negative samples, thereby mitigating the optimization difficulties when using the BCE loss only and resulting in improved classification ability. To validate our perspective, we conduct theoretical analysis and extensive empirical evaluations on public datasets. Additionally, we successfully integrate the ranking loss into Tencent's online advertising system, achieving notable lifts of 0.70% and 1.26% in Gross Merchandise Value (GMV) for two main scenarios. The code is openly accessible at: https://github.com/SkylerLinn/Understanding-the-Ranking-Loss.

We present Tencent's ads recommendation system and examine the challenges and practices of learning appropriate recommendation representations. Our study begins by showcasing our approaches to preserving prior knowledge when encoding features of diverse types into embedding representations. We specifically address sequence features, numeric features, and pre-trained embedding features. Subsequently, we delve into two crucial challenges related to feature representation: the dimensional collapse of embeddings and the interest entanglement across different tasks or scenarios. We propose several practical approaches to address these challenges that result in robust and disentangled recommendation representations. We then explore several training techniques to facilitate model optimization, reduce bias, and enhance exploration. Additionally, we introduce three analysis tools that enable us to study feature correlation, dimensional collapse, and interest entanglement. This work builds upon the continuous efforts of Tencent's ads recommendation team over the past decade. It summarizes general design principles and presents a series of readily applicable solutions and analysis tools. The reported performance is based on our online advertising platform, which handles hundreds of billions of requests daily and serves millions of ads to billions of users.

Click-through rate (CTR) prediction plays a pivotal role in real-world applications, particularly in recommendation systems and online advertising. A significant research branch in this domain focuses on user behavior modeling. Current research predominantly centers on modeling co-occurrence relationships between the target item and items previously interacted with by users. However, this focus neglects the intricate modeling of user behavior patterns. In reality, the abundance of user interaction records encompasses diverse behavior patterns, indicative of a spectrum of habitual paradigms. These patterns harbor substantial potential to significantly enhance CTR prediction performance. To harness the informational potential within behavior patterns, we extend Target Attention (TA) to Target Pattern Attention (TPA) to model pattern-level dependencies. Furthermore, three critical challenges demand attention: the inclusion of unrelated items within patterns, data sparsity of patterns, and computational complexity arising from numerous patterns. To address these challenges, we introduce the Deep Pattern Network (DPN), designed to comprehensively leverage information from behavior patterns. DPN efficiently retrieves target-related behavior patterns using a target-aware attention mechanism. Additionally, it contributes to refining patterns through a pre-training paradigm based on self-supervised learning while promoting dependency learning within sparse patterns. Our comprehensive experiments, conducted across three public datasets, substantiate the superior performance and broad compatibility of DPN.

Side-information integrated sequential recommendation incorporates supplementary information to alleviate the issue of data sparsity. The state-of-the-art works mainly leverage some side information to improve the attention calculation to learn user representation more accurately. However, there are still some limitations to be addressed in this topic. Most of them merely learn the user representation at the item level and overlook the association of the item sequence and the side-information sequences when calculating the attentions, which results in the incomprehensive learning of user representation. Some of them learn the user representations at both the item and side-information levels, but they still face the problem of insufficient optimization of multiple user representations. To address these limitations, we propose a novel model, i.e., Multi-Sequence Sequential Recommender (MSSR), which learns the user's multiple representations from diverse sequences. Specifically, we design a multi-sequence integrated attention layer to learn more attentive pairs than the existing works and adaptively fuse these pairs to learn user representation. Moreover, our user representation alignment module constructs the self-supervised signals to optimize the representations. Subsequently, they are further refined by our side information predictor during training. For item prediction, our MSSR extra considers the side information of the candidate item, enabling a comprehensive measurement of the user's preferences. Extensive experiments on four public datasets show that our MSSR outperforms eleven state-of-the-art baselines. Visualization and case study also demonstrate the rationality and interpretability of our MSSR.

Multi-task learning (MTL) models have demonstrated impressive results in computer vision, natural language processing, and recommender systems. Even though many approaches have been proposed, how well these approaches balance different tasks on each parameter still remains unclear. In this paper, we propose to measure the task dominance degree of a parameter by the total updates of each task on this parameter. Specifically, we compute the total updates by the exponentially decaying Average of the squared Updates (AU) on a parameter from the corresponding task. Based on this novel metric, we observe that many parameters in existing MTL methods, especially those in the higher shared layers, are still dominated by one or several tasks. The dominance of AU is mainly due to the dominance of accumulative gradients from one or several tasks. Motivated by this, we propose a Task-wise Adaptive learning rate approach, AdaTask in short, to separate the accumulative gradients and hence the learning rate of each task for each parameter in adaptive learning rate approaches (e.g., AdaGrad, RMSProp, and Adam). Comprehensive experiments on computer vision and recommender system MTL datasets demonstrate that AdaTask significantly improves the performance of dominated tasks, resulting SOTA average task-wise performance. Analysis on both synthetic and real-world datasets shows AdaTask balance parameters in every shared layer well.

Auxiliary-Task Learning (ATL) aims to improve the performance of the target task by leveraging the knowledge obtained from related tasks. Occasionally, learning multiple tasks simultaneously results in lower accuracy than learning only the target task, which is known as negative transfer. This problem is often attributed to the gradient conflicts among tasks, and is frequently tackled by coordinating the task gradients in previous works. However, these optimization-based methods largely overlook the auxiliary-target generalization capability. To better understand the root cause of negative transfer, we experimentally investigate it from both optimization and generalization perspectives. Based on our findings, we introduce ForkMerge, a novel approach that periodically forks the model into multiple branches, automatically searches the varying task weights by minimizing target validation errors, and dynamically merges all branches to filter out detrimental task-parameter updates. On a series of auxiliary-task learning benchmarks, ForkMerge outperforms existing methods and effectively mitigates negative transfer.

Multi-task learning (MTL) has been widely used in recommender systems, wherein predicting each type of user feedback on items (e.g, click, purchase) are treated as individual tasks and jointly trained with a unified model. Our key observation is that the prediction results of each task may contain task-specific knowledge about user’s fine-grained preference towards items. While such knowledge could be transferred to benefit other tasks, it is being overlooked under the current MTL paradigm. This paper, instead, proposes a Cross-Task Knowledge Distillation framework that attempts to leverage prediction results of one task as supervised signals to teach another task. However, integrating MTL and KD in a proper manner is non-trivial due to several challenges including task conflicts, inconsistent magnitude and requirement of synchronous optimization. As countermeasures, we 1) introduce auxiliary tasks with quadruplet loss functions to capture cross-task fine-grained ranking information and avoid task conflicts, 2) design a calibrated distillation approach to align and distill knowledge from auxiliary tasks, and 3) propose a novel error correction mechanism to enable and facilitate synchronous training of teacher and student models. Comprehensive experiments are conducted to verify the effectiveness of our framework in real-world datasets.

In Click-through rate (CTR) prediction models, a user’s interest is usually represented as a fixed-length vector based on her history behaviors. Recently, several methods are proposed to learn an attentive weight for each user behavior and conduct weighted sum pooling. However, these methods only manually select several fields from the target item side as the query to interact with the behaviors, neglecting the other target item fields, as well as user and context fields. Directly including all these fields in the attention may introduce noise and deteriorate the performance. In this paper, we propose a novel model named AutoAttention, which includes all item/user/context side fields as the query, and assigns a learnable weight for each field pair between behavior fields and query fields. Pruning on these field pairs via these learnable weights lead to automatic field pair selection, so as to identify and remove noisy field pairs. Though including more fields, the computation cost of AutoAttention is still low due to using a simple attention function and field pair selection. Extensive experiments on the public dataset and Tencent’s production dataset demonstrate the effectiveness of the proposed approach.

Collaborative filtering (CF), as a standard method for recommendation with implicit feedback, tackles a semi-supervised learning problem where most interaction data are unobserved. Such a nature makes existing approaches highly rely on mining negatives for providing correct training signals. However, mining proper negatives is not a free lunch, encountering with a tricky trade-off between mining informative hard negatives and avoiding false ones. We devise a new approach named as Hardness-Aware Debiased Contrastive Collaborative Filtering (HDCCF) to resolve the dilemma. It could sufficiently explore hard negatives from two-fold aspects: 1) adaptively sharpening the gradients of harder instances through a set-wise objective, and 2) implicitly leveraging item/user frequency information with a new sampling strategy. To circumvent false negatives, we develop a principled approach to improve the reliability of negative instances and prove that the objective is an unbiased estimation of sampling from the true negative distribution. Extensive experiments demonstrate the superiority of the proposed model over existing CF models and hard negative mining methods.

In online display advertising, guaranteed contracts and real-time bidding (RTB) are two major ways to sell impressions for a publisher. For large publishers, simultaneously selling impressions through both guaranteed contracts and in-house RTB has become a popular choice. Generally speaking, a publisher needs to derive an impression allocation strategy between guaranteed contracts and RTB to maximize its overall outcome (e.g., revenue and/or impression quality). However, deriving the optimal strategy is not a trivial task, e.g., the strategy should encourage incentive compatibility in RTB and tackle common challenges in real-world applications such as unstable traffic patterns (e.g., impression volume and bid landscape changing). In this paper, we formulate impression allocation as an auction problem where each guaranteed contract submits virtual bids for individual impressions. With this formulation, we derive the optimal bidding functions for the guaranteed contracts, which result in the optimal impression allocation. In order to address the unstable traffic pattern challenge and achieve the optimal overall outcome, we propose a multi-agent reinforcement learning method to adjust the bids from each guaranteed contract, which is simple, converging efficiently and scalable. The experiments conducted on real-world datasets demonstrate the effectiveness of our method.

In online display advertising, advertisers usually participate in real-time bidding to acquire ad impression opportunities. In most advertising platforms, a typical impression acquiring demand of advertisers is to maximize the sum value of winning impressions under budget and some key performance indicators constraints, (e.g. maximizing clicks with the constraints of budget and cost per click upper bound). The demand can be various in value type (e.g. ad exposure/click), constraint type (e.g. cost per unit value) and constraint number. Existing works usually focus on a specific demand or hardly achieve the optimum. In this paper, we formulate the demand as a constrained bidding problem, and deduce a unified optimal bidding function on behalf of an advertiser. The optimal bidding function facilitates an advertiser calculating bids for all impressions with only m parameters, where m is the constraint number. However, in real application, it is non-trivial to determine the parameters due to the non-stationary auction environment. We further propose a reinforcement learning (RL) method to dynamically adjust parameters to achieve the optimum, whose converging efficiency is significantly boosted by the recursive optimization property in our formulation. We name the formulation and the RL method, together, as Unified Solution to Constrained Bidding (USCB). USCB is verified to be effective on industrial datasets and is deployed in Alibaba display advertising platform.

Since 2019, most ad exchanges and sell-side platforms (SSPs), in the online advertising industry, shifted from second to first price auctions. Due to the fundamental difference between these auctions, demand-side platforms (DSPs) have had to update their bidding strategies to avoid bidding unnecessarily high and hence overpaying. Bid shading was proposed to adjust the bid price intended for second-price auctions, in order to balance cost and winning probability in a first-price auction setup. In this study, we introduce a novel deep distribution network for optimal bidding in both open (non-censored) and closed (censored) online first-price auctions. Offline and online A/B testing results show that our algorithm outperforms previous state-of-art algorithms in terms of both surplus and effective cost per action (eCPX) metrics. Furthermore, the algorithm is optimized in run-time and has been deployed into VerizonMedia DSP as production algorithm, serving hundreds of billions of bid requests per day. Online A/B test shows that advertiser's ROI are improved by +2.4%, +2.4%, and +8.6% for impression based (CPM), click based (CPC), and conversion based (CPA) campaigns respectively.

The delayed feedback problem is one of the imperative challenges in online advertising, which is caused by the highly diversified feedback delay of a conversion varying from a few minutes to several days. It is hard to design an appropriate online learning system under these non-identical delay for different types of ads and users. In this paper, we propose to tackle the delayed feedback problem in online advertising by "Following the Prophet" (FTP for short). The key insight is that, if the feedback came instantly for all the logged samples, we could get a model without delayed feedback, namely the "prophet". Although the prophet cannot be obtained during online learning, we show that we could predict the prophet's predictions by an aggregation policy on top of a set of multi-task predictions, where each task captures the feedback patterns of different periods. We propose the objective and optimization approach for the policy, and use the logged data to imitate the prophet. Extensive experiments on three real-world advertising datasets show that our method outperforms the previous state-of-the-art baselines.

Click-through rate (CTR) prediction is a crucial task in recommender systems and online advertising. The embedding-based neural networks have been proposed to learn both explicit feature interactions through a shallow component and deep feature interactions by a deep neural network (DNN) component. These sophisticated models, however, slow down the prediction inference by at least hundreds of times. To address the issue of significantly increased serving latency and high memory usage for real-time serving in production, this paper presents DeepLight: a framework to accelerate the CTR predictions in three aspects: 1) accelerate the model inference via explicitly searching informative feature interactions in the shallow component; 2) prune redundant parameters at the inter-layer level in the DNN component; 3) prune the dense embedding vectors to make them sparse in the embedding matrix. By combining the above efforts, the proposed approach accelerates the model inference by 46X on Criteo dataset and 27X on Avazu dataset without any loss on the prediction accuracy. This paves the way for successfully deploying complicated embedding-based neural networks in real-world serving systems.

In e-commerce advertising, the ad platform usually relies on auction mechanisms to optimize different performance metrics, such as user experience, advertiser utility, and platform revenue. However, most of the state-of-the-art auction mechanisms only focus on optimizing a single performance metric, e.g., either social welfare or revenue, and are not suitable for e-commerce advertising with various, dynamic, difficult to estimate, and even conflicting performance metrics. In this paper, we propose a new mechanism called Deep GSP auction, which leverages deep learning to design new rank score functions within the celebrated GSP auction framework. These new rank score functions are implemented via deep neural network models under the constraints of monotone allocation and smooth transition. The requirement of monotone allocation ensures Deep GSP auction nice game theoretical properties, while the requirement of smooth transition guarantees the advertiser utilities would not fluctuate too much when the auction mechanism switches among candidate mechanisms to achieve different optimization objectives. We deployed the proposed mechanisms in a leading e-commerce ad platform and conducted comprehensive experimental evaluations with both offline simulations and online A/B tests. The results demonstrated the effectiveness of the Deep GSP auction compared to the state-of-the-art auction mechanisms.

Click-through rate (CTR) prediction plays a critical role in recommender systems and online advertising. The data used in these applications are multi-field categorical data, where each feature belongs to one field. Field information is proved to be important and there are several works considering fields in their models. In this paper, we proposed a novel approach to model the field information effectively and efficiently. The proposed approach is a direct improvement of FwFM, and is named Field-matrixed Factorization Machines (FmFM, or FM2). We also proposed a new explanation of FM and FwFM within the FmFM framework, and compared it with the FFM. Besides pruning the cross terms, our model supports field-specific variable dimensions of embedding vectors, which acts as a soft pruning. We also proposed an efficient way to minimize the dimension while keeping the model performance. The FmFM model can also be optimized further by caching the intermediate vectors, and it only takes thousands of floating-point operations (FLOPs) to make a prediction. Our experiment results show that it can out-perform the FFM, which is more complex. The FmFM model’s performance is also comparable to DNN models which require much more FLOPs in runtime.

Online auctions play a central role in online advertising, and are one of the main reasons for the industry's scalability and growth. With great changes in how auctions are being organized, such as changing the second- to first-price auction type, advertisers and demand platforms are compelled to adapt to a new volatile environment. Bid shading is a known technique for preventing overpaying in auction systems that can help maintain the strategy equilibrium in first-price auctions, tackling one of its greatest drawbacks. In this study, we propose a machine learning approach of modeling optimal bid shading for non-censored online first-price ad auctions. We clearly motivate the approach and extensively evaluate it in both offline and online settings on a major demand side platform. The results demonstrate the superiority and robustness of the new approach as compared to the existing approaches across a range of performance metrics.

Conversion prediction plays an important role in online advertis- ing since Cost-Per-Action (CPA) has become one of the primary campaign performance objectives in the industry. Unlike click pre- diction, conversions have different types in nature, and each type may be associated with different decisive factors. In this paper, we formulate conversion prediction as a multi-task learning problem, so that the prediction models for different types of conversions can be learned together. These models share feature representa- tions, but have their specific parameters, providing the benefit of information-sharing across all tasks. We then propose Multi-Task Field-weighted Factorization Machine (MT-FwFM) to solve these tasks jointly. Our experiment results show that, compared with two state-of-the-art models, MT-FwFM improve the AUC by 0.74% and 0.84% on two types of conversions, and the weighted AUC across all conversion types is also improved by 0.50%.

Optimizing news headlines is important for publishers and media sites. A compelling headline will increase readership, user engagement and social shares. At Yahoo Front Page, headline testing is carried out using a test-rollout strategy: we first allocate equal proportion of the traffic to each headline variation for a defined testing period, and then shift all future traffic to the best-performing variation. In this paper, we introduce a multi-armed bandit (MAB) approach with batched Thompson Sampling (bTS) to dynamically test headlines for news articles. This method is able to gradually allocate traffic towards optimal headlines while testing. We evaluate the bTS method based on empirical impressions/clicks data and simulated user responses. The result shows that the bTS method is robust, converges accurately and quickly to the optimal headline, and outperforms the test-rollout strategy by 3.69% in terms of clicks.

Click-through rate (CTR) prediction is a critical task in online display advertising. The data involved in CTR prediction are typically multi-field categorical data, i.e., every feature is categorical and belongs to one and only one field. One of the interesting characteristics of such data is that features from one field often interact differently with features from different other fields. Recently, Field-aware Factorization Machines (FFMs) have been among the best performing models for CTR prediction by explicitly modeling such difference. However, the number of parameters in FFMs is in the order of feature number times field number, which is unacceptable in the real-world production systems. In this paper, we propose Field-weighted Factorization Machines (FwFMs) to model the different feature interactions between different fields in a much more memory-efficient way. Our experimental evaluations show that FwFMs can achieve competitive prediction performance with only as few as 4% parameters of FFMs. When using the same number of parameters, FwFMs can bring 0.92% and 0.47% AUC lift over FFMs on two real CTR prediction data sets.