Recommender Systems

On a delivery platform, personalized store ranking greatly influences what users find and order. Unlike digital-only domains, candidate stores are local and bound by real-time availability and delivery operations. One central modeling tension is between surfacing new stores for trial and preserving ranking quality for sessions with reorder intent. We present Universal Venue Ranker (UVR), a production system deployed at Wolt that pairs a bidirectional transformer encoder for sequential user modeling with a GBDT ranker integrating contextual, user, and store features. Trained across all stores and domains of a country while enforcing local delivery constraints at inference, UVR replaces four previously separate ranking models (three for restaurants, one for retail) with a single unified system. Label smoothing and trial-biased sample weighting steer the model toward new stores, lifting offline trial MRR by +12% to +30% over production while regressing reorder MRR in five of six countries. These regressions leave Global CVR, our core online metric, which blends trial and reorder sessions, statistically unchanged. We validate UVR in three consecutive A/B tests, the first two across Wolt's largest operating markets and the third spanning all operating countries and both domains. UVR V1 delivers +5.5% Merchant Trial Rate and +0.16% Global CVR over the previous production ranker; V2 adds a further +0.45% Merchant Trial Rate on top; and V3, our cross-domain unification of the restaurant and retail rankers, adds a further +1.31% Retail Merchant Trial Rate, together accounting for substantial incremental gross order value and a materially simplified serving stack.

Generative recommendation has emerged as a promising end-to-end paradigm for personalized recommendation. However, user preferences continuously evolve over time, making self-evolving an essential capability for generative recommender systems. Existing evolving strategies, such as continual retraining and distillation-based adaptation, directly update the shared model parameters using streaming interactions. Nevertheless, we find that directly applying such strategies to generative recommendation introduces a critical issue, termed evolution conflict. Specifically, heterogeneous preference shifts from different users are optimized within a fully shared autoregressive parameter space, causing dominant behavioral patterns to progressively dominate the model evolution process while underrepresented patterns become increasingly overlooked. To address this issue, we propose a self-evolving memory paradigm for generative recommendation, aiming to enable effective evolution across heterogeneous behavioral patterns. We further identify three key principles for effective self-evolving recommendation systems, including isolated memorization, reinforced evolution, and scalable application. Guided by these principles, we develop LION, a simple yet effective framework centered on a sparse Key-Value memory layer. Specifically, LION introduces sparse memory activation to isolate the evolution of different behavioral patterns, while a consolidation loss is designed to reinforce the learning of underrepresented preference dynamics during continual adaptation. Extensive experiments on diverse real-world datasets demonstrate the effectiveness of LION under various continual evolution settings (e.g., per-period evaluation, user/item group evaluation, and evolution convergence analysis). The codes are released at https://github.com/JazyJiang/Self-Evolving-Memory-for-Generative-Recommendation.

First-stage recommenders in multi-stage systems produce a ranked candidate list from which a limited prefix is forwarded to downstream rankers. Because each forwarded item must be processed by more expensive ranking stages, this shortlist cannot be arbitrarily large. The first-stage objective is therefore high coverage of relevant items within the forwarded prefix, commonly measured by Recall@$k$. A relevant item may be available deeper in the retrieved list but absent from the shorter prefix that is actually consumed. This paper studies post-hoc verification for promoting such candidates into the consumed shortlist without retraining or replacing the retriever. We introduce a lightweight generative verifier for retrieval models. Given a retriever state and a candidate item, the verifier scores the item through the likelihood of its identifier tokens. It is trained post hoc with next-token cross entropy, requires no sampled negatives or candidate pool during training, and scores only the retriever's top-$K$ candidates at inference. The interface is minimal: the retriever supplies a query state and candidate items, and the item representation can use any fixed tokenization. Across Amazon product recommendation and YaMBDa music recommendation, the same verifier training recipe improves Recall@10 for SASRec, GRU4Rec, NextItNet, and MiniOneRec. Ablations show that the improvements are not explained solely by injecting item-content features into the retriever, supporting verification as a post-hoc output-side adaptation mechanism.

Recently, a wide range of recommendation algorithms inspired by deep learning techniques have emerged as the performance leaders on several standard recommendation benchmarks. While these algorithms were built on different DL techniques (e.g., dropouts, autoencoder), they have similar performance and even similar cost functions. This paper studies whether the models' comparable performance are sheer coincidence, or they can be unified under a single framework. We find that all linear performance leaders effectively add only a nuclear-norm based regularizer, or a Frobenius-norm based regularizer. The former ones possess a (surprising) rigid structure that limits the models' predictive power but their solutions are low rank and have closed form. The latter ones are more expressive and more efficient for recommendation but their solutions are either full-rank or require executing hard-to-tune numeric procedures such as ADMM. Along this line of finding, we further propose two low-rank, closed-form solutions, derived from carefully generalizing Frobenius-norm based regularizers. The new solutions get the best of both nuclear-norm and Frobenius-norm world.

Auto-research agents have shown the potential to automate hypothesis generation, experiment execution, and iterative refinement. However, scaling this paradigm to industry-scale recommendation models introduces two challenges: (1) long feedback loops, where model training can take days, making serial iteration prohibitively slow and requiring parallel exploration across multiple research directions; and (2) system complexity, where large configurations, fragile infrastructure dependencies, and multi-day GPU jobs require robust and recoverable execution. We present Auto-RecSys, an autonomous research system for long-horizon experimentation on industry-scale recommendation models. Auto-RecSys addresses these challenges through three harness designs: (1) distributed asynchronous execution for running multiple experiments in parallel across servers, (2) centralized cross-server memory for persistent and recoverable execution across sessions and failures, and (3) cognitive-procedural separation, where natural-language skill files guide LLM reasoning while deterministic scripts enforce operational correctness. Auto-RecSys further employs a dual-loop self-evolving architecture: an Execution Evolution Loop in which model-specific playbooks accumulate operational knowledge by recording failed attempts and crystallizing successful pipelines, and an Idea Evolution Loop in which experimental outcomes inform subsequent ideation. Evaluated on recommendation models, Auto-RecSys significantly reduces the human time required per experiment cycle and improves execution reliability as its playbooks mature.

Recommender systems have become core infrastructure for modern online platforms, personalizing content at scale and strongly influencing what users see, click on, and purchase. However, this dependence on user interaction also exposes them to shilling attacks, where malicious actors can inject fake profiles to distort item rankings and control visibility. Existing attacks often rely on target-specific fine-tuning or fixed profile templates, making them either difficult to adapt to different victims or easier to detect. To overcome these limitations, we propose the Agentic Group Attack System (AGAS), a coordinated shilling framework where a central Coordinator directs a group of role-switching worker agents to adaptively promote a target item across different victim families. The Coordinator dynamically adjusts the strategy when progress stalls or suppression signals increase, while workers pursue a shared objective and switch between active and inactive roles to avoid repetitive patterns. Under the same attack budgets and evaluation protocols, AGAS consistently surpasses strong baselines in target promotion while better preserving benign recommendation quality, weakening representative detectors, and achieving higher efficiency than prior attacks. These findings also emphasize that defending recommender systems may require mechanisms that can handle adaptive shilling campaigns, not just isolated fake-profile injections. Our code is available at https://github.com/phkhanhtrinh23/AGAS.

Accurate citations are the foundation of academic writing, tracing intellectual origins and substantiating core claims. However, manually navigating the growing volume of scientific literature is increasingly difficult, prompting reliance on automatic citation recommendation. While modern retrieval-augmented architectures have largely mitigated the fabrication of non-existent papers, current systems relying on semantic similarity struggle with misattribution, often citing authentic papers that fail to logically support the author's claim. To address this challenge, we argue that accurate citation requires a shift from similarity-based search to active, claim-level reasoning. We propose ReCite, a decoupled agentic framework that orchestrates location perception, intent-aware query planning, and reflective verification. Trained on synthesized reasoning trajectories, our agent verifies claim-evidence consistency and triggers self-correction loops when retrieved candidates lack logical support. Experiments demonstrate that our lightweight framework outperforms state-of-the-art massive generative models in strict citation accuracy. By grounding literature matching in verifiable logic rather than semantic overlap, ReCite establishes a reliable foundation for automated academic writing.

Short-form video platforms use recommender systems to maximize engagement through highly efficient personalized recommendations. However, the impact of these recommendations on users with ADHD compared to users without ADHD remains underexplored. Through this study, we introduce and operationalize the Engagement Trap, illustrating how recommender systems, while successfully optimizing for engagement, disproportionately disadvantage users with ADHD. This stratified study of 302 participants, recruited via the online platform Prolific, compares experiences between participants with and without ADHD. Our results show that while recommendations are perceived as relevant across groups, participants with ADHD report significantly higher levels of time blindness, post-usage regret, and emotional distress when consuming recommendations. Moreover, we collect feedback for several proof-of- concept, theoretical design interventions for neuro-inclusive design principles. These findings provide quantitative evidence of systemic differences in engagement-optimized recommender systems and highlight the unbalanced negative effects and interactions these systems create for participants with ADHD. We argue for neurodiversity-aware, human-centered design approaches that mitigate such algorithmic harms and support more equitable experiences.

This paper studies adaptive recommendation under intent drift, where feedback from each recommendation outcome can reveal whether the relational evidence used for ranking is useful, missing, or misleading. While Knowledge Graphs (KGs) provide essential semantic structure to handle these shifts, traditional KG-enhanced systems treat the graph as a static retrieval substrate, making it brittle to evolving intents, noisy metadata, and recurring failure patterns. This paper proposes CGM-Rec, a continual graph memory framework for adaptive recommendation. CGM-Rec treats the graph state as a writable memory and maintains two complementary components. Therein, a Semantic Graph Memory is updated conservatively through quality-gated typed operations for storing stable and high-confidence relational knowledge. Meanwhile, an Episodic Lesson Memory acts as a fast reactive memory that learns recent outcomes, failure cases, and corrective hints. During testing, model parameters remain frozen and adaptation occurs only through memory writes. We evaluate CGM-Rec under a frozen-parameter, one-pass reranking protocol, where encoders and prompts remain fixed during testing and adaptation occurs only through memory writes. Experiments across multiple recommendation settings show that CGM-Rec improves over evaluated neural and LLM-based baselines on most metrics. Particularly, under sampled-candidate reranking, CGM-Rec improves HR@1 by up to 29.58% over the strongest LLM baseline on Bundle, and outperforms K-RagRec on metadata-rich ML-100K with HR@5 of 0.5941 versus 0.4746.

Production recommender systems shape what billions of people see, and sustaining their performance requires continual optimization: as content, user behavior, and upstream models shift, the choices governing retrieval, ranking, and serving must be revisited. Traditionally, human engineers test such changes through online experiments--a slow, reactive process limited by engineering effort, leaving parts of the system unrevised as conditions change. Although large language models have been applied to ranking, user modeling, and offline model development, few systems place an agent in a continual closed loop that acts on a live recommender and learns from the measured effects of its decisions. We present CORAL (Constraint-Optimized Recommender via an Agentic Loop), an LLM-native harness that closes this loop: each cycle, the agent observes operating signals, reasons over a memory of past decisions and outcomes, and invokes tools--including a numerical optimizer that keeps changes within a fixed operating budget--to reconfigure the recommender, with measured outcomes informing the next cycle. We formulate this as a partially observed, non-stationary, constrained optimization problem in which the policy improves in context, without parameter updates, from its prior actions. Across two large-scale social platforms, evaluated with A/B experiments, the same harness improves engagement at no additional serving cost on one and reduces serving cost without degrading engagement on the other, spanning the engagement-efficiency frontier. Performance improves as the loop iterates, suggesting that a single agentic loop can automate continual optimization work traditionally performed by human algorithm engineers under explicit guardrails.

Recommender-system experiments often rely on a single random training seed, assuming that run-to-run stochasticity has limited impact on evaluation conclusions. This assumption is risky, as a training seed may influence several algorithm-dependent mechanisms, including parameter initialization, mini-batch ordering, dropout, masking, latent sampling, and training-time negative sampling. We examine this assumption by fixing the data partition and varying the training seed across hyperparameter configurations. We analyze seed effects at three levels: user-level metric sensitivity, validation-based model selection and recommendation-list agreement. Results show that seed variation is often detectable. Its impact depends on whether configurations are clearly separated, whether validation results transfer to test, and whether similar scores lead to similar top-$k$ lists. Findings suggest that reporting single-seed results can overstate the stability of recommender system evaluation, and that training seeds should be treated as part of the evaluation protocol rather than as incidental implementation noise.

Advertising recommendation requires continuously tuning complex system parameters while balancing commercial returns and user experience. Recent work has introduced large language models (LLMs) with skill documents to assist this labor-intensive process, but skill optimization remains largely prompt-driven, lacking a principled mechanism to attribute rewards to specific document edits. To address this limitation, we propose Document-Mediated Reinforcement Learning (DMRL), a skill self-evolution framework that models skill document optimization as a sequence of structured editing actions. In DMRL, an upper-level agent performs controlled document edits, while a frozen lower-level task agent evaluates their effects through A/B testing. To address credit assignment and long-term outcomes, we introduce two key components: (1) Dual-Relative Policy Optimization (DRPO), a post-training policy optimization method for robust and risk-aware advantage estimation; and (2) Long-term Reward Predictor (LRP), which estimates long-term outcomes by modeling population heterogeneity with disentangled representation learning and cross-attention transfer. DMRL was deployed on a large-scale short-video ads platform and extensive empirical evaluation shows that DMRL outperforms state-of-the-art baselines across key advertising metrics

Serving useful recommendations under distribution shift is crucial for balancing utility and risk in out-of-distribution (OOD) recommendation. However, most existing OOD methods improve ranking or construct counterfactual candidates without controlling the proxy-label false discovery rate (FDR) of the served set. In this work, we formulate OOD serving as the $α$-Valid Counterfactual Recommendation ($α$-VCR) problem to retain candidate support learned from counterfactual supervision while controlling proxy-label FDR, and propose GenCAR, which couples preference-grounded counterfactual supervision with calibrated set selection. In particular, GenCAR fixes the stable-preference representation while intervening on the environmental factor, grounds offline large language model proposals through preference anchors and trust-radius filtering, and uses conformal $p$-values for Benjamini--Hochberg selection. We theoretically bound conditional counterfactual approximation error and prove finite-sample, distribution-free control of proxy-label FDR under exchangeability and positive regression dependence, with a Benjamini--Yekutieli guarantee under arbitrary dependence. Extensive experiments audit realized proxy false discovery proportions and demonstrate that GenCAR consistently enhances OOD candidate recovery across diverse benchmarks.

Multimodal Recommender Systems (MRSs) typically rely on a flawed "modality harmony" assumption, presuming that multimodal features are inherently beneficial and strictly aligned with users' collaborative interaction patterns. However, modality-topology conflicts are ubiquitous in real-world scenarios due to deceptive visual clickbaits and mismatched semantics. Blindly integrating these noisy modalities inevitably pollutes the pristine collaborative space, causing severe representation distortion. To address this, we propose Orthogonal purification and topology-guided MoE for conflict-aware multimodal Recommendation (OrthoRec). At its core, OrthoRec introduces Collaborative-Guided Orthogonal Purification (CGOP), which geometrically decouples multimodal features into directions parallel and orthogonal to a pure collaborative anchor. By adaptively truncating the orthogonal noise with an energy-preserving normalization, CGOP rectifies deceptive semantic directions while preserving the modality's intrinsic representation capacity. Furthermore, we design a Topology-Aware Routing Mixture-of-Experts (TAR-MoE). Guided by the collaborative topology, TAR-MoE employs decoupled sigmoid gating to break the zero-sum bottleneck of traditional softmax attention, autonomously determining the injection scale for each purified modality. Finally, a safe-SSL objective is introduced to dynamically penalize the forced contrastive alignment of contradictory pairs. Experiments on three real-world Amazon datasets show that OrthoRec consistently outperforms competitive recent baselines and exhibits improved robustness under modality noise and item sparsity.

Scaling Transformers has driven large gains in language modeling, but transplanting this to behavior-sequence modeling in production ranking is challenging: recommendation differs in signal quality, where behavior sequences are noisy, temporally irregular, and sparsely supervised, and in computation asymmetry, where each request scores many candidates against one shared user history under tight latency budgets. We propose ReST, a recommendation-native Transformer scaling framework. For signal quality, it introduces a sequence encoder with dual-gated attention, rotary positional and temporal embedding, stabilized residual normalization, and training-only auxiliary objectives. For computation asymmetry, it factorizes ranking into a heavy reusable encoder and a lightweight cross decoder with projection-free KV attention and token-specific parameterization, coupling user-level shared-prefix training with shared-prefix serving for compute-once, decode-many-times ranking. Across industrial and public benchmarks, ReST achieves higher accuracy and scales more consistently along sequence length, depth, and width, where LLM-style Transformer blocks saturate. A one-week online A/B test on a production advertising platform improves online AUC by 1.31% and lifts a core revenue metric by 11.93% within a 50 ms P99 budget; ReST has since been fully deployed in production, showing that behavior-sequence scaling remains a promising, under-exploited axis for production ranking.

User representation pre-training has become a fundamental paradigm for alleviating data sparsity in downstream personalization tasks. However, existing studies predominantly focus on single-app or app-level behaviors, overlooking the inherently cross-source and multi-granular nature of user activities on mobile devices. At the device level, user intent emerges from complex interactions among heterogeneous behavior sources and hierarchical action structures, posing challenges that cannot be addressed by conventional app-centric modeling. To tackle this issue, we propose CM-PTM, a novel Cross Multi-source Behavior Pre-Training Model tailored for mobile game user representation learning on device-level behavioral logs. CM-PTM employs hierarchical cascaded mask-then-predict proxy tasks that first infer the source of the next behavior and then progressively refine predictions at the app-action level. This design enables unified modeling of cross-source dependencies and fine-grained behavioral dynamics within a single pre-training paradigm. Extensive experiments on large-scale real-world mobile datasets demonstrate that CM-PTM effectively captures users' endogenous interests and consistently delivers significant performance gains on downstream mobile game recommendation tasks.

We investigate the role of conversational context modeling in user preference tracking for Conversational Recommendation Systems (CRSs). In this regard, we propose DREAMS, a novel tree-structured context modeling framework that explicitly captures user preference evolution throughout multi-turn interactions. DREAMS introduces two specialized node types to support the two fundamental objectives of CRSs: preference elicitation and preference exploitation. Specifically, elicitation nodes leverage Monte Carlo Tree Search (MCTS) to strategically explore conversational actions and infer latent user preferences, while exploitation nodes employ LLM-based refinement to transform the tracked preference state into structured retrieval queries for recommendation. Extensive experiments on benchmark datasets demonstrate the effectiveness of DREAMS and its design.

Large language models (LLMs) are increasingly used as rerankers in conversational recommender systems, yet measured gains depend strongly on the retrieval and inference protocol. On the ReDial conversational movie recommendation benchmark, we compare proprietary, open-weight, and fine-tuned LLM rerankers with collaborative-filtering and sequential baselines in a shared retrieve-then-rerank pipeline. We vary candidate-pool size, first-stage retriever, and decoding temperature. With a shared semantic top-250 candidate pool and strict candidate-aware scoring, the best proprietary reranker reaches NDCG@10 of 0.1497, compared with 0.0939 for the strongest non-LLM baseline. The same reranker reaches 0.2925 in zero-shot generation, showing that unconstrained scoring can yield a much larger apparent advantage than matched-pool evaluation. No evaluated open-weight LLM outperforms the tuned shallow autoencoder baseline under this protocol. For the strongest proprietary and open-weight rerankers, switching from semantic to collaborative-filtering candidates raises NDCG@10 by more than 50%, showing that measured reranker performance is highly sensitive to candidate generation. For the best proprietary reranker, raising temperature from 0 to 1.0 increases top-10 Jaccard distance from 0.0900 to 0.1240 while mean NDCG@10 changes negligibly, whereas weaker LLMs show larger degradation. These ReDial results support treating candidate generation, candidate-pool size, scoring policy, and decoding configuration as required reporting fields rather than implementation details.

Next-basket repurchase recommendation is commonly formulated as a ranking task: given a customer's purchase history, the system ranks previously purchased items that may be needed again. In production settings, however, ranking accuracy is only one component of recommendation quality. Customers may also benefit from concise evidence about why an item is recommended now. Large language models (LLMs) offer a potential way to surface such evidence through feature-based, human-readable rationales grounded in interpretable behavioral signals. We construct repurchase features spanning cadence, frequency, recency, user behavior, and item popularity, and evaluate LLMs on two public grocery datasets and one proprietary retail dataset. We investigate (1) whether off-the-shelf LLMs can use these features as next-basket scorers relative to heuristic and supervised rankers, and (2) whether LLM-cited features carry outcome-grounded ranking signal. For the latter, we compare LLM-cited features with model-specific attribution methods under a cross-model feature-masking protocol that measures ranking degradation after masking selected features. Our results show that LLM scores are not competitive with supervised rankers, suggesting that off-the-shelf LLMs should not be used as standalone repurchase recommenders. However, changes in prompt and evidence representation can improve outcome-grounded feature-masking results in some settings even when ranking performance does not improve; the effect is dataset-dependent and does not consistently match attribution baselines. These findings suggest a practical role for LLMs as validated explanation components rather than primary rankers, with rationale quality evaluated separately from ranking accuracy.

Repurchase recommenders in e-commerce are commonly framed as a binary question asking "will this customer buy this item within W days", a formulation that requires a separately trained model for every horizon of interest. We replace this stack with survival models that predict time-to-repurchase directly, and evaluate them on millions of customers from a major grocery e-commerce platform across more than thirty ablation configurations. Our study makes three contributions. First, an empirical hazard analysis reveals a slightly decreasing marginal hazard (k ~ 0.9), differing from the common intuition that grocery items become more likely to be repurchased the longer since the last purchase (increasing hazard, k > 1). Log-Normal achieves the best marginal fit (R^2 = 0.998) and the best ranking, despite Weibull providing the best conditional residual fit, revealing an apparent discrepancy we analyze in detail. Second, a single Accelerated Failure Time (AFT) model replaces three per-horizon binary classifiers, matching or exceeding each at its own horizon while using roughly 3x fewer total trees. Feature importance reshuffles under the survival objective: channel-cadence and recency signals rise while aggregate frequency counts fall. Third, a 4-parameter parametric calibration maps raw survival CDFs to per-horizon probabilities with zero cross-horizon monotonicity violations. Calibration quality varies by an order of magnitude across the AFT family: Exponential AFT (Weibull k=1) achieves expected calibration error (ECE) ~1e-4, roughly 10x lower than Log-Normal, while ranking metrics agree within 0.3% relative. We adopt Exponential AFT for probability-consuming surfaces and Log-Normal for pure ranking, exposing a principled calibration-ranking trade-off within a single AFT family.

Music recommendation relies primarily on two signals: user-item interactions, which fail in the cold-start regime, and intrinsic musical content, available for any recording. We argue that a third, largely untapped signal is both richer and more principled: critical adjacency, the pairwise relation established when an expert critic explicitly links two artists in long-form prose. It encodes deliberate judgments about which artists belong together. Prior work established its internal validity, showing it recovers coherent, interpretable communities and can match collaborative filtering in user-satisfaction simulations, with no user data. What has been missing is external validation: whether this critic-sourced relation is grounded in the music itself versus sociological context. We test it against acoustic content, reframing the question as one of construct validity. Representing artists as empirical distributions over 80 low-level Essentia acoustic descriptors and modeling pairwise proximity via marginal optimal-transport (Wasserstein) distances, we evaluate how far critical adjacency is sonically recoverable under a cold-start, artist-disjoint split. Our ensemble recovers these edges at out-of-sample AUC of 0.767 (95% CI 0.761-0.775). Recoverability rises monotonically with critical consensus, reaching 0.865 on multi-source attested edges. Stratified evaluations align with sociological models of genre: tightly bounded, scene-based genres show higher recoverability than broad industry umbrella terms. Critical discourse is thus a rich source of information for recommendation, decomposing into a reproducible "sonic core" and a "sociological remainder" driven by narrative positioning, subcultural context, and canonical placement. The work offers both a scalable cold-start discovery mechanism and a sociologically grounded approach to MIR and MRS research.

Semantic identifiers (SIDs) represent entities as hierarchical token sequences for generative retrieval and recommendation. Residual-quantization tokenizers construct these sequences by selecting a codeword at each level and passing a residual to the next. We view this process as progressive commonality removal: each token captures a component shared within its group, while later tokens should model the remaining differences. This view reveals three limitations: a corpus-wide shared component can consume first-level capacity, hard assignment ignores graded similarities to nearby codewords, and full-codeword subtraction can leave variation along the selected-codeword direction in the next residual. We therefore develop our solution in the post-hoc setting, where residual construction is not constrained by input reconstruction. Specifically, we propose PRQ-KMeans, which removes the global-mean component, refines centroids with Top-k similarity-weighted updates, and replaces full-codeword subtraction with a projection residual that removes each representation's selected-centroid component. Experiments on a large-scale industrial search dataset and four public recommendation benchmarks show that PRQ-KMeans achieves the strongest overall performance among the evaluated tokenizers, including gains of up to 7.4% in HitRate and 11.8% in MRR on the industrial dataset.

LLM-as-a-Judge, which leverages a large language model to evaluate natural language generated by another AI application or model, has become a standard, scalable approach for accelerating and extending costly human evaluation. Yet most work treats a judge as a static artifact, evaluating it once at construction or against a fixed benchmark. We argue instead that an LLM judge operating in a deployed system is better understood as having a lifecycle. It must be built, trained, deployed, and continuously maintained as the surrounding data evolves, and each phase poses distinct technical and operational challenges. We present such a lifecycle for the LLM judges that evaluate recommendation explanations at Netflix. Everything we report comes out of a series of controlled online member-facing experiments, in which our pipeline generated and the judges assessed hundreds of thousands of distinct show-level explanations per week across a changing catalog. Our framework has four phases. (I) Birth defines the evaluation criteria and builds curated benchmark datasets with human labels and rationales. (II) Training refines the judges' rubrics via Reasoning-Aligned Rubric Tuning (RART), which uses a meta-judge over reasoning output as the learning signal. (III) Deployment puts one judge in two online roles, quality gating and reflective generation. (IV) Monitoring runs a continuous Human-in-the-Loop (HITL) alignment process that detects drift and triggers re-tuning behind a human review gate. We report results from a five-week online A/B test over tens of millions of members on the Netflix mobile app, in which judge-aligned explanations shifted member viewing toward novel content (previously unwatched) and increased successful browse-to-play sessions relative to a no-explanation control, with no quality-related escalations.

Search and recommendation serve a shared discovery objective but encode intent differently. We study this boundary through Dear Algo on Threads, a deployed product where open-ended requests such as \emph{more NBA news} or \emph{less politics} steer subsequent feed recommendations rather than return a one-shot result list. Its agentic intent layer compiles explicit, inferred, negative, and compound intent into a grounded executable plan, then invokes conventional retrieval and optional semantic or multimodal reranking. The layer shares an intent-to-retrieval contract without requiring one model or serving path across search-like and recommendation-like modes. We evaluate Dear Algo under a precision-first objective. In a blinded audit of 300 public request-item pairs (296 evaluable), a strict categorical LLM-as-a-judge gate achieved 94.4% exact-Relevant precision [88.8%, 98.9%]. Across 72 normalized request clusters, the full configuration produced 7.73 judge-qualified candidates per 20 slots versus 6.61 for an LLM-derived-query baseline, a gain of 1.11 [0.12, 2.12]. In a candidate-randomized serving-path study restricted to the reranker path's first 72 eligible hours, the user-weighted judge-Irrelevant share among judged admissions was 2.80% versus 4.78% off (-1.97 points [-3.02, -0.94]), while Exact-Relevant share was 2.24 points higher [0.08, 4.41]. Together, these studies show how explicit natural-language intent can be carried into feed recommendation under a precision-first evaluation framework

Semantic-ID generative recommenders represent each item as a short sequence of discrete semantic tokens and predict the next item by autoregressively generating this token sequence. This paradigm enables a unified generation interface for item IDs, histories, and item text, but it also creates a structured optimization bottleneck during reward-based post-training: when an early semantic token enters the wrong branch of the item-token space, finite rollout groups rarely reach the ground-truth item, so group-relative optimization receives identical zero rewards and produces no useful advantage. We propose Hint-Conditioned Generative Recommendation (HCGRec), a semantic-ID generative recommendation framework that recovers learning signal for such hard training instances. HCGRec diagnoses each instance with checkpoint rollouts and supplies a minimal target-prefix hint only when the current generator cannot reach the correct item. The model then generates the unhinted suffix under the hinted semantic branch, turning zero-reward groups into informative comparisons over item-token completions. Hinting also changes token identity: hinted prefix tokens are oracle-provided item context, while unhinted suffix tokens are sampled generation actions. We therefore introduce hint-aware credit decomposition, using supervised learning to preserve item-semantic and prefix-structure alignment for hinted tokens and GRPO to optimize the sampled suffix. Experiments on sequential recommendation benchmarks show that HCGRec substantially improves over supervised fine-tuning and vanilla reward-based post-training, while reducing zero-advantage training samples from over 70% to below 20%. The code is accessible at https://github.com/WncFht/GRec.

Traditional offline recommendation evaluation relies heavily on complex, manually maintained feature pipelines that are difficult to scale. While Large Language Models (LLMs) offer a promising alternative by predicting user engagement directly from raw text logs, empirical analysis in this study identifies a critical failure mode termed bidirectional rationalization. In a zero-shot setting, LLMs are found to convincingly argue for both positive and negative user engagement outcomes on the exact same item with identical evidence, highlighting the unreliability of off-the-shelf LLMs in predicting user engagement. To resolve this, we develop and apply a sequential behavioral alignment framework pairing fine-tuning with preference optimization over paired correct and counterfactual rationales. Evaluated on real-world homepage interaction logs, this aligned reasoning approach achieves a 32.19% lift in Macro-F1 score over the zero-shot baseline and matches the production feature-engineered baseline. The results demonstrate that behavioral alignment mitigates bidirectional rationalization while delivering human-interpretable reasoning traces without manual pipeline overhead.

Modern recommender systems treat observed actions as reliable proxies for user preferences, yet interactions often reflect exploration or comparison rather than stable preference expression. As interfaces evolve from static layouts toward generative UIs and immersive extended reality (XR), the need for deeper, modality-agnostic user understanding grows: these adaptive environments must decide not only what to present but where, when, how prominently, and most importantly why a user acts. We propose an Inverse Theory of Mind (IToM) pipeline that reasons backward from observed interactions to infer the beliefs, preferences, and decision-making traits that explain behavior. The pipeline reconstructs each user's decision context, including what was chosen and what alternatives were available, applies LLM-driven counterfactual reasoning to produce evidence-grounded natural-language belief statements, and synthesizes these beliefs through multi-hypothesis abductive inference into a structured user persona. We evaluate on the OPeRA dataset against ground-truth personality assessments, attitudinal surveys, and interview-based personas across four tasks: next action prediction, shopping attitude alignment, Big Five personality inference, and held-out category prediction. Results show that inferred personas match or exceed ground-truth personas and that multi-hypothesis reasoning is essential for accurate personality prediction. We further demonstrate cross-modal transferability with a persona-driven spatial banking application on VisionOS.

Large-scale neural recommender systems are typically trained with a softmax cross-entropy objective over the full item vocabulary. For a typical large number of possible items $K$, the final classification layer dominates memory, requiring $O(nK)$ logits and gradients to materialize for a batch of $n$ examples. Sampled softmax reduces this cost by restricting the objective to only $k \ll K$ candidate negative items, resulting in an $O(nk)$ memory. However, for a fixed budget $B = n k$, it remains unclear whether one should prioritize larger batches or the inclusion of more negative items. We address this question by analyzing sampled-softmax training under a fixed memory constraint. Under standard smoothness and variance assumptions, our theoretical evidence suggests that the fastest convergence arises from an $n \sim B, k \sim 1$ allocation. So, an actionable rule is to include as many objects as possible given computational constraints. Our theory is supported by controlled synthetic and synthetic and four real sequential recommendation benchmarks, including MovieLens-20M. The suggested configuration achieve faster convergence and better final recommendation quality than imbalanced alternatives within the same memory constraint. These findings provide a theoretical and empirical foundation for configuring memory during the training of recommender systems. Code, reproducibility materials, and all scripts for generating figures are available at https://anonymous.4open.science/r/LimitedMemoryRule-BBFB

Multi-modal recommenders fuse user-item interaction signals with item modalities such as text, images, and audio, but the usefulness of each drifts over time and at different rates. For example, around Valentine's Day, chocolate purchases become less driven by textual ingredient cues and more by visual packaging and ambient audio. This \emph{modality time-scale mismatch} gives rise to two coupled challenges: (1) users with different temporal behavior profiles require different modality proportions, and (2) less relevant modalities are more likely to introduce outdated or misleading signals into the recommender. We address both challenges within a unified diffusion-based recommender, \textbf{TimeRoute}. A temporal-aware modal router maps each user's aggregated temporal profile to a personalized modality distribution, replacing the globally shared fusion weights used in prior work. The diffusion-based graph reconstructor is conditioned on the same profile through Feature-wise Linear Modulation (FiLM) with dual-stream long- and short-term denoising heads. This design captures both slowly and rapidly evolving temporal dynamics to suppress outdated modality edges before they enter the propagation graph. Experiments on TikTok, Amazon-Baby, and Amazon-Sports, averaged over 10 seeds, demonstrate consistent improvements over strong baselines across Recall@K, Precision@K, and NDCG@K, reaching up to 9.8% (P@20 on Amazon-Baby). Controlled attribution studies further show that these gains require both the proposed mechanisms and temporal input: naively granting the backbone the same temporal profile yields no benefit, and feeding the router random noise performs no better than removing the router entirely. Code is available at https://anonymous.4open.science/r/TimeRoute.

Cross-domain recommendation (CDR) transfers preference knowledge across related domains, but federated deployment makes cross-domain alignment difficult because the behavioral anchors that align item spaces, such as overlapping users and shared interaction signals, are often sparse, unavailable, or privacy-sensitive across clients. To address this tension, we revisit federated CDR as generation over a stable semantic item language. By representing items as discrete semantic ID (SID) sequences derived from public item-side metadata, cross-domain item alignment is induced by a shared vocabulary rather than by exchanging private interactions or aligning domain-specific embeddings. Directly federating SID-based generators, however, introduces two design constraints: the SID tokenizer must remain fixed to preserve cross-client token consistency, which creates a semantic-only bottleneck because local collaborative filtering (CF) signals cannot be globally shared or aligned; meanwhile, standard federated averaging can cause negative transfer under domain heterogeneity. To overcome these constraints, we propose FedCGR, a federated generative CDR framework that keeps the item language stable and makes adaptation explicit. FedCGR injects local CF evidence through a reliability-aware semantic interface and trains a prototype-personalized generator that selectively aggregates shared parameters according to domain relatedness while keeping domain-specific quantities local. Experiments on six Amazon cross-domain scenarios show that FedCGR consistently outperforms federated generative baselines and achieves competitive performance against strong sequential and federated CDR methods under both full-ranking and sampled evaluation protocols.