While Large Language Models (LLMs) excel at many reasoning tasks, their native inability to produce calibrated, multi-class probability distributions limits their use in high-stakes Web applications like content moderation and fraud detection. Existing methods to elicit probabilities from LLMs either sacrifice their crucial Chain-of-Thought (CoT) reasoning capabilities or suffer from poor calibration. To address this, we introduce a new paradigm, Verbalized Probability Distribution, and a novel training framework, RLVP (Reinforcement Learning with Verbalized Probabilities). RLVP fine-tunes an LLM to generate both an interpretable CoT and a complete, verbalized probability distribution. We overcome the ''insufficient reward granularity'' problem in standard Reinforcement Learning (RL) for classification by using soft probabilities from expert tabular models as a dense reward curriculum. Through large-scale joint training on 169 tabular tasks, we demonstrate that a single RLVP-trained model can surpass a strong, task-specific XGBoost baseline on up to 55% of tasks. More importantly, the trained model achieves state-of-the-art few-shot performance on unseen, heterogeneous Web benchmarks that mix structured data with free text, achieving performance comparable to or superior than expert models trained on the same limited data. This showcases a strong capability for generalization and knowledge transfer to complex Web data. Our work presents a viable path toward building general-purpose, probabilistically-sound, and interpretable foundation models for the Web.

Predictive modeling on web-scale tabular data presents significant scalability challenges for industrial applications, often involving billions of instances and hundreds of heterogeneous numerical features. The inherent complexities of these features—characterized by anisotropy, heavy-tailed distributions, and non-stationarity—not only impose bottlenecks on the training efficiency and scalability of mainstream models like Gradient Boosting Decision Trees (GBDTs), but also compel practitioners into laborious, inefficient, and expert-dependent manual feature engineering. To systematically address this challenge, we introduce KMLP, a novel hybrid deep architecture. KMLP synergistically integrates a shallow Kolmogorov-Arnold Network (KAN) as a front-end with a Gated Multilayer Perceptron (gMLP) as the backbone. The KAN front-end leverages its learnable activation functions to automatically model complex non-linear transformations for each input feature in an end-to-end manner, thereby automating feature representation learning. Subsequently, the gMLP backbone efficiently captures high-order interactions among these refined representations. Extensive experiments on multiple public benchmarks and an ultra-large-scale industrial web dataset with billions of samples demonstrate that KMLP achieves state-of-the-art (SOTA) performance. Crucially, our findings reveal that KMLP's performance advantage over strong baselines like GBDTs becomes more pronounced as the data scale increases. This validates KMLP as a scalable and adaptive deep learning paradigm, offering a promising path forward for modeling large-scale, dynamic web tabular data.

A critical limitation of conventional sequential recommendation models (SRMs) is their reliance on observed user-item interaction sequences within a closed-world setting, which hinders their ability to generalize to unseen or infrequent items. Recently, Large Language Models (LLMs) have shown remarkable promise in recommendation systems due to their vast world knowledge and advanced reasoning capabilities. Current research has predominantly explored two approaches: using LLMs to directly generate recommendations and distilling knowledge from LLMs to enhance conventional SRMs. However, these approaches face two major challenges: (1) high inference costs, as they require LLM responses during inference, either for generating predictions or as supplementary input; (2) inadequate distillation of the reasoning process, as existing methods focus mainly on improving embeddings or aligning outputs, without fully integrating LLMs' inherent reasoning capabilities. To address these issues, we propose LCKD-SR, anLLM-drivenCascadedKnowledgeDistillation framework forSequentialRecommendation. In this framework, an LLM, a Teacher SRM, and a Student SRM form a hierarchical distillation structure, enabling an LLM-free inference by using only the Student model. Beyond traditional embedding and ranking distillation, our framework abstracts the LLM's sequential reasoning abilities by identifying key interactions that subsequently guide the Teacher's attention using learnable markers. The Student model, which mirrors the architecture of the Teacher, achieves seamless knowledge alignment from the Teacher across all three aspects. Extensive experiments demonstrate the effectiveness and efficiency of the proposed LCKD-SR, showcasing its scalability to perform multi-level knowledge transfer while enabling LLM-independent inference, thereby overcoming the inference cost and reasoning limitations of existing methods.

Anti-fraud machine learning systems are perpetually confronted with the significant challenge of concept drift, driven by the continuous and intense evolution of fraudulent techniques. That is, outdated models trained on historical fraudulent behaviors often fall short in addressing the evolving tactics of malicious users over time. The key issue lies in effectively tackling the rapid and significant evolution of fraudsters' behaviors to detect these emerging and unforeseen anomalies. In this paper, we propose a solution by directly accessing real-time data and introducing a lightweight plug-in approach named TRE (Test-time Retrieval-based Representation Enrichment). Considering the similarity among samples, TRE employs a retriever to efficiently identify the top-K most relevant recent samples and implements an aggregation strategy to provide neighboring embeddings to the predictor. It thus adjusts the trained classifiers during the test time, providing them with the information from the latest unlabeled data. Extensive experiments on three large-scale real-world datasets demonstrate the superiority of TRE. By consistently incorporating information from the nearest neighbors, TRE demonstrates high adaptability and surpasses existing methods in performance.

Accurate personalized headline generation hinges on precisely capturing user interests from historical behaviors. However, existing methods neglect personalized-irrelevant click noise in entire historical clickstreams, which may lead to hallucinated headlines that deviate from genuine user preferences. In this paper, we reveal the detrimental impact of click noise on personalized generation quality through rigorous analysis in both user and news dimensions. Based on these insights, we propose a novel Personalized Headline Generation framework via Denoising Fake Interests from Implicit Feedback (PHG-DIF). PHG-DIF first employs dual-stage filtering to effectively remove clickstream noise, identified by short dwell times and abnormal click bursts, and then leverages multi-level temporal fusion to dynamically model users' evolving and multi-faceted interests for precise profiling. Moreover, we release DT-PENS, a new benchmark dataset comprising the click behavior of 1,000 carefully curated users and nearly 10,000 annotated personalized headlines with historical dwell time annotations. Extensive experiments demonstrate that PHG-DIF substantially mitigates the adverse effects of click noise and significantly improves headline quality, achieving state-of-the-art (SOTA) results on DT-PENS. Our framework implementation and dataset are available at https://github.com/liukejin-up/PHG-DIF.

To migrate the remarkable successes of Large Language Models (LLMs), the community has made numerous efforts to generalize them to the table reasoning tasks for the widely deployed tabular data. Despite that, in this work, by showing a probing experiment on our proposed StructQA benchmark, we postulate that even the most advanced LLMs (such as GPTs) may still fall short of coping with tabular data. More specifically, the current scheme often simply relies on serializing the tabular data, together with the meta information, then inputting them through the LLMs. We argue that the loss of structural information is the root of this shortcoming. In this work, we further propose TAMO, which bears an ideology to treat the tables as an independent modality integrated with the text tokens. The resulting model in TAMO is a multimodal framework consisting of a hypergraph neural network as the global table encoder seamlessly integrated with the mainstream LLM. Empirical results on various benchmarking datasets, including HiTab, WikiTQ, WikiSQL, FeTaQA, and StructQA, have demonstrated significant improvements on generalization with an average relative gain of **42.65%**.

Deep learning-based classifiers have been widely used in the field of financial fraud transaction detection. However, training a high-performance classifier for fraud detection is challenging due to the lack of sufficient labeled fraud data. Particularly, it is difficult to detect stealthy fraud transactions that closely mimic benign user behaviors. We observe that the suspicious transactions identified by the online detection system can augment the feature space to improve the detection performance of machine learning-based models. In this paper, we propose a new framework GIANTESS to leverage suspicious transactions to augment the feature space and thus enhance the detection of stealthy fraud transactions. Our semi-supervised approach combines both labeled transactions and unlabeled suspicious transactions to train a detection model. Specifically, it first estimates pseudo labels of suspicious transactions and then combines the pseudo labels with ground truth labels to train the detection model. We conduct experiments on two real-world datasets to demonstrate the effectiveness of our proposed method on detecting stealthy fraud transactions. The experimental results show that GIANTESS successfully improves the recall by up to 6.3% at the fixed low false positive rate of 1%. We also perform a 9-week deployment test of our system in a real-world online payment platform to demonstrate the performance of GIANTESS.

In many practical binary classification applications, such as financial fraud detection or medical diagnosis, it is crucial to optimize a model's performance on high-confidence samples whose scores are higher than a specific threshold, which is calculated by a given false positive rate according to practical requirements. However, the proportion of high-confidence samples is typically extremely small, especially in long-tailed datasets, which can lead to poor recall results and an alignment bias between realistic goals and loss. To address this challenge, we propose a novel loss reweighting framework called Momentum Threshold-Oriented Loss (MTOL) for binary classification tasks and propose two instantiated losses of it. Given a limited FPR range, MTOL aims to improve the recall of binary classification models at that FPR range by incorporating a batch memory queue and momentum estimation mechanisms. The MTOL adaptively estimates thresholds of FPR during the model training iterations and up-weights the loss of samples in the threshold range, with little consumption of storage and computation. Our experimental results on various datasets, including CIFAR-10, CIFAR-100, Tiny-ImageNet, demonstrate the significant effect of MTOL in improving the recall at low FPR especially in class imbalance settings. These results suggest that MTOL is a promising approach in scenarios where the model's performance in the low FPR range is of utmost importance.

With the availability of gene expression data by RNA-seq, powerful statistical approaches for grouping similar gene expression profiles across different environments have become increasingly important. We describe and assess a computational model for clustering genes into distinct groups based on the pattern of gene expression in response to changing environment. The model capitalizes on the Poisson distribution to capture the count property of RNA-seq data. A two-stage hierarchical expectation–maximization (EM) algorithm is implemented to estimate an optimal number of groups and mean expression amounts of each group across two environments. A procedure is formulated to test whether and how a given group shows a plastic response to environmental changes. The impact of gene–environment interactions on the phenotypic plasticity of the organism can also be visualized and characterized. The model was used to analyse an RNA-seq dataset measured from two cell lines of breast cancer that respond differently to an anti-cancer drug, from which genes associated with the resistance and sensitivity of the cell lines are identified. We performed simulation studies to validate the statistical behaviour of the model. The model provides a useful tool for clustering gene expression data by RNA-seq, facilitating our understanding of gene functions and networks.

The formation of phenotypic traits, such as biomass production, tumor volume and viral abundance, undergoes a complex process in which interactions between genes and developmental stimuli take place at each level of biological organization from cells to organisms. Traditional studies emphasize the impact of genes by directly linking DNA-based markers with static phenotypic values. Functional mapping, derived to detect genes that control developmental processes using growth equations, has proven powerful for addressing questions about the roles of genes in development. By treating phenotypic formation as a cohesive system using differential equations, a different approach-systems mapping-dissects the system into interconnected elements and then map genes that determine a web of interactions among these elements, facilitating our understanding of the genetic machineries for phenotypic development. Here, we argue that genetic mapping can play a more important role in studying the genotype-phenotype relationship by filling the gaps in the biochemical and regulatory process from DNA to end-point phenotype. We describe a new framework, named network mapping, to study the genetic architecture of complex traits by integrating the regulatory networks that cause a high-order phenotype. Network mapping makes use of a system of differential equations to quantify the rule by which transcriptional, proteomic and metabolomic components interact with each other to organize into a functional whole. The synthesis of functional mapping, systems mapping and network mapping provides a novel avenue to decipher a comprehensive picture of the genetic landscape of complex phenotypes that underlie economically and biomedically important traits.

Because of its widespread occurrence and role in shaping evolutionary processes in the biological kingdom, especially in plants, polyploidy has been increasingly studied from cytological to molecular levels. By inferring gene order, gene distances and gene homology, linkage mapping with molecular markers has proven powerful for investigating genome structure and organization. Here we review and assess a general statistical model for three-point linkage analysis in autotetraploids by integrating double reduction, a phenomenon that commonly occurs in autopolyploids whose chromosomes are derived from a single ancestral species. This model does not require any assumption on the distribution of the occurrence of double reduction and can handle the complexity of multilocus linkage in terms of crossover interference. Implemented with the expectation-maximization (EM) algorithms, the model can estimate and test the recombination fractions between less informative dominant markers, thus facilitating its practical implications for any autopolyploids in most of which inexpensive dominant markers are still used for their genetic and evolutionary studies. The model was applied to reanalyze a published data in tetraploid switchgrass, validating its practical usefulness and utilization.

Despite our expanding knowledge about the biochemistry of gene regulation involved in host-pathogen interactions, a quantitative understanding of this process at a transcriptional level is still limited. We devise and assess a computational framework that can address this question. This framework is founded on a mixture model-based likelihood, equipped with functionality to cluster genes per dynamic and functional changes of gene expression within an interconnected system composed of the host and pathogen. If genes from the host and pathogen are clustered in the same group due to a similar pattern of dynamic profiles, they are likely to be reciprocally co-evolving. If genes from the two organisms are clustered in different groups, this means that they experience strong host-pathogen interactions. The framework can test the rates of change for individual gene clusters during pathogenic infection and quantify their impacts on host-pathogen interactions. The framework was validated by a pathological study of poplar leaves infected by fungal Marssonina brunnea in which co-evolving and interactive genes that determine poplar-fungus interactions are identified. The new framework should find its wide application to studying host-pathogen interactions for any other interconnected systems.