SORT: A Systematically Optimized Ranking Transformer for Industrial-scale Recommenders

While Transformers have achieved remarkable success in LLMs through superior scalability, their application in industrial-scale ranking models remains nascent, hindered by the challenges of high feature sparsity and low label density. In this paper, we propose SORT (Systematically Optimized Ranking Transformer), a scalable model designed to bridge the gap between Transformers and industrial-scale ranking models. We address the high feature sparsity and low label density challenges through a series of optimizations, including request-centric sample organization, local attention, query pruning and generative pre-training. Furthermore, we introduce a suite of refinements to the tokenization, multi-head attention (MHA), and feed-forward network (FFN) modules, which collectively stabilize the training process and enlarge the model capacity. To maximize hardware efficiency, we optimize our training system to elevate the model FLOPs utilization (MFU) to 22%. Extensive experiments demonstrate that SORT outperforms strong baselines and exhibits excellent scalability across data size, model size and sequence length, while remaining flexible at integrating diverse features. Finally, online A/B testing in large-scale e-commerce scenarios confirms that SORT achieves significant gains in key business metrics, including orders (+6.35%), buyers (+5.97%) and GMV (+5.47%), while simultaneously halving latency (-44.67%) and doubling throughput (+121.33%).

ProductResearch: Training E-Commerce Deep Research Agents via Multi-Agent Synthetic Trajectory Distillation

Large Language Model (LLM)-based agents show promise for e-commerce conversational shopping, yet existing implementations lack the interaction depth and contextual breadth required for complex product research. Meanwhile, the Deep Research paradigm, despite advancing information synthesis in web search, suffers from domain gaps when transferred to e-commerce. We propose ProductResearch, a multi-agent framework that synthesizes high-fidelity, long-horizon tool-use trajectories for training robust e-commerce shopping agents. The framework employs a User Agent to infer nuanced shopping intents from behavioral histories, and a Supervisor Agent that orchestrates iterative collaboration with a Research Agent to generate synthetic trajectories culminating in comprehensive, insightful product research reports. These trajectories are rigorously filtered and distilled through a reflective internalization process that consolidates multi-agent supervisory interactions into coherent single-role training examples, enabling effective fine-tuning of LLM agents for complex shopping inquiries. Extensive experiments show that a compact MoE model fine-tuned on our synthetic data achieves substantial improvements over its base model in response comprehensiveness, research depth, and user-perceived utility, approaching the performance of frontier proprietary deep research systems and establishing multi-agent synthetic trajectory training as an effective and scalable paradigm for enhancing LLM-based shopping assistance.

Learning to Reflect and Correct: Towards Better Decoding Trajectories for Large-Scale Generative Recommendation

Generative Recommendation (GR) has become a promising paradigm for large-scale recommendation systems. However, existing GR models typically perform single-pass decoding without explicit refinement, causing early deviations to accumulate and ultimately degrade recommendation quality. To tackle this problem, we propose GRC, which is, to our knowledge, the first structured reflection-correction framework for GR that extends standard decoding into a Generation-Reflection-Correction (GRC) process. Concretely, GRC introduces a supervised reflection-correction template that decomposes the decoding process into initial draft generation, multi-granular reflection, and reflection-guided correction, thereby enabling structured reflection and correction in the semantic token space. To further explore the enlarged refinement space introduced by the GRC process, we optimize the entire GRC trajectory with GRPO-based reinforcement learning, under a carefully designed reward function with token-level and trajectory-level signals. For efficient online serving, we propose an Entropy-Guided Reflection Scheduling (EGRS) strategy that dynamically allocates more correction budget to high-uncertainty decoding trajectories during beam search. Extensive experiments on real-world datasets show that GRC consistently outperforms six state-of-the-art baselines by up to 15.74%, and online A/B tests demonstrate its substantial practical value in large-scale industrial recommendation, delivering a 1.79% lift in advertising revenue with only modest latency overhead.

SIGMA: A Semantic-Grounded Instruction-Driven Generative Multi-Task Recommender at AliExpress

With the rapid evolution of Large Language Models, generative recommendation is gradually reshaping the paradigm of recommender systems. However, most existing methods are still confined to the interaction-driven next-item prediction paradigm, failing to rapidly adapt to evolving trends or address diverse recommendation tasks along with business-specific requirements in real-world scenarios. To this end, we present SIGMA, a Semantic-Grounded Instruction-Driven Generative Multi-Task Recommender at AliExpress. Specifically, we first ground item entities in general semantics via a unified latent space capturing both semantic and collaborative relations. Building upon this, we develop a hybrid item tokenization method for precise modeling and efficient generation. Moreover, we construct a large-scale multi-task SFT dataset to empower SIGMA to fulfill various recommendation demands via instruction-following. Finally, we design a three-step item generation procedure integrated with an adaptive probabilistic fusion mechanism to calibrate the output distributions based on task-specific requirements for recommendation accuracy and diversity. Extensive offline experiments and online A/B tests demonstrate the effectiveness of SIGMA.

Constraint-Aware Generative Auto-bidding via Pareto-Prioritized Regret Optimization

Auto-bidding systems aim to maximize marketing value while satisfying strict efficiency constraints such as Target Cost-Per-Action (CPA). Although Decision Transformers provide powerful sequence modeling capabilities, applying them to this constrained setting encounters two challenges: 1) standard Return-to-Go conditioning causes state aliasing by neglecting the cost dimension, preventing precise resource pacing; and 2) standard regression forces the policy to mimic average historical behaviors, thereby limiting the capacity to optimize performance toward the constraint boundary. To address these challenges, we propose PRO-Bid, a constraint-aware generative auto-bidding framework based on two synergistic mechanisms: 1) Constraint-Decoupled Pareto Representation (CDPR) decomposes global constraints into recursive cost and value contexts to restore resource perception, while reweighting trajectories based on the Pareto frontier to focus on high-efficiency data; and 2) Counterfactual Regret Optimization (CRO) facilitates active improvement by utilizing a global outcome predictor to identify superior counterfactual actions. By treating these high-utility outcomes as weighted regression targets, the model transcends historical averages to approach the optimal constraint boundary. Extensive experiments on two public benchmarks and online A/B tests demonstrate that PRO-Bid achieves superior constraint satisfaction and value acquisition compared to state-of-the-art baselines.

Thinking Broad, Acting Fast: Latent Reasoning Distillation from Multi-Perspective Chain-of-Thought for E-Commerce Relevance

Effective relevance modeling is crucial for e-commerce search, as it aligns search results with user intent and enhances customer experience. Recent work has leveraged large language models (LLMs) to address the limitations of traditional relevance models, especially for long-tail and ambiguous queries. By incorporating Chain-of-Thought (CoT) reasoning, these approaches improve both accuracy and interpretability through multi-step reasoning. However, two key limitations remain: (1) most existing approaches rely on single-perspective CoT reasoning, which fails to capture the multifaceted nature of e-commerce relevance (e.g., user intent vs. attribute-level matching vs. business-specific rules); and (2) although CoT-enhanced LLM's offer rich reasoning capabilities, their high inference latency necessitates knowledge distillation for real-time deployment, yet current distillation methods discard the CoT rationale structure at inference, using it as a transient auxiliary signal and forfeiting its reasoning utility. To address these challenges, we propose a novel framework that better exploits CoT semantics throughout the optimization pipeline. Specifically, the teacher model leverages Multi-Perspective CoT (MPCoT) to generate diverse rationales and combines Supervised Fine-Tuning (SFT) with Direct Preference Optimization (DPO) to construct a more robust reasoner. For distillation, we introduce Latent Reasoning Knowledge Distillation (LRKD), which endows a student model with a lightweight inference-time latent reasoning extractor, allowing efficient and low-latency internalization of the LLM's sophisticated reasoning capabilities. Evaluated in offline experiments and online A/B tests on an e-commerce search advertising platform serving tens of millions of users daily, our method delivers significant offline gains, showing clear benefits in both commercial performance and user experience.

Masked Diffusion Generative Recommendation

Generative recommendation (GR) typically first quantizes continuous item embeddings into multi-level semantic IDs (SIDs), and then generates the next item via autoregressive decoding. Although existing methods are already competitive in terms of recommendation performance, directly inheriting the autoregressive decoding paradigm from language models still suffers from three key limitations: (1) autoregressive decoding struggles to jointly capture global dependencies among the multi-dimensional features associated with different positions of SID; (2) using a unified, fixed decoding path for the same item implicitly assumes that all users attend to item attributes in the same order; (3) autoregressive decoding is inefficient at inference time and struggles to meet real-time requirements. To tackle these challenges, we propose MDGR, a Masked Diffusion Generative Recommendation framework that reshapes the GR pipeline from three perspectives: codebook, training, and inference. (1) We adopt a parallel codebook to provide a structural foundation for diffusion-based GR. (2) During training, we adaptively construct masking supervision signals along both the temporal and sample dimensions. (3) During inference, we develop a warm-up-based two-stage parallel decoding strategy for efficient generation of SIDs. Extensive experiments on multiple public and industrial-scale datasets show that MDGR outperforms ten state-of-the-art baselines by up to 10.78%. Furthermore, by deploying MDGR on a large-scale online advertising platform, we achieve a 1.20% increase in revenue, demonstrating its practical value. The code will be released upon acceptance.

Infer As You Train: A Symmetric Paradigm of Masked Generative for Click-Through Rate Prediction

Generative models are increasingly being explored in click-through rate (CTR) prediction field to overcome the limitations of the conventional discriminative paradigm, which rely on a simple binary classification objective. However, existing generative models typically confine the generative paradigm to the training phase, primarily for representation learning. During online inference, they revert to a standard discriminative paradigm, failing to leverage their powerful generative capabilities to further improve prediction accuracy. This fundamental asymmetry between the training and inference phases prevents the generative paradigm from realizing its full potential. To address this limitation, we propose the Symmetric Masked Generative Paradigm for CTR prediction (SGCTR), a novel framework that establishes symmetry between the training and inference phases. Specifically, after acquiring generative capabilities by learning feature dependencies during training, SGCTR applies the generative capabilities during online inference to iteratively redefine the features of input samples, which mitigates the impact of noisy features and enhances prediction accuracy. Extensive experiments validate the superiority of SGCTR, demonstrating that applying the generative paradigm symmetrically across both training and inference significantly unlocks its power in CTR prediction.

Synergistic Integration and Discrepancy Resolution of Contextualized Knowledge for Personalized Recommendation

The integration of large language models (LLMs) into recommendation systems has revealed promising potential through their capacity to extract world knowledge for enhanced reasoning capabilities. However, current methodologies that adopt static schema-based prompting mechanisms encounter significant limitations: (1) they employ universal template structures that neglect the multi-faceted nature of user preference diversity; (2) they implement superficial alignment between semantic knowledge representations and behavioral feature spaces without achieving comprehensive latent space integration. To address these challenges, we introduce CoCo, an end-to-end framework that dynamically constructs user-specific contextual knowledge embeddings through a dual-mechanism approach. Our method realizes profound integration of semantic and behavioral latent dimensions via adaptive knowledge fusion and contradiction resolution modules. Experimental evaluations across diverse benchmark datasets and an enterprise-level e-commerce platform demonstrate CoCo's superiority, achieving a maximum 8.58% improvement over seven cutting-edge methods in recommendation accuracy. The framework's deployment on a production advertising system resulted in a 1.91% sales growth, validating its practical effectiveness. With its modular design and model-agnostic architecture, CoCo provides a versatile solution for next-generation recommendation systems requiring both knowledge-enhanced reasoning and personalized adaptation.

CSMF: Cascaded Selective Mask Fine-Tuning for Multi-Objective Embedding-Based Retrieval

Multi-objective embedding-based retrieval (EBR) has become increasingly critical due to the growing complexity of user behaviors and commercial objectives. While traditional approaches often suffer from data sparsity and limited information sharing between objectives, recent methods utilizing a shared network alongside dedicated sub-networks for each objective partially address these limitations. However, such methods significantly increase the model parameters, leading to an increased retrieval latency and a limited ability to model causal relationships between objectives. To address these challenges, we propose the Cascaded Selective Mask Fine-Tuning (CSMF), a novel method that enhances both retrieval efficiency and serving performance for multi-objective EBR. The CSMF framework selectively masks model parameters to free up independent learning space for each objective, leveraging the cascading relationships between objectives during the sequential fine-tuning. Without increasing network parameters or online retrieval overhead, CSMF computes a linearly weighted fusion score for multiple objective probabilities while supporting flexible adjustment of each objective's weight across various recommendation scenarios. Experimental results on real-world datasets demonstrate the superior performance of CSMF, and online experiments validate its significant practical value.