Retrieval-augmented generation (RAG) enhances LLMs with external knowledge, yet generation remains vulnerable to retrieval-induced noise and uncertain placement of relevant chunks, often causing hallucinations. We present Ext2Gen, an extract-then-generate framework that strengthens LLMs via joint evidence selection and answer generation, dynamically identifying query-relevant content while suppressing noise, thereby removing the need for any independent pre-generation compression module. Optimized through preference alignment with well-curated pairwise feedback, Ext2Gen produces accurate and faithful answers even under noisy or imprecise retrieval. Experiments demonstrate that it substantially enhances the robustness of the generation backbone and yields greater performance gains than methods relying on independent compression models, (e.g., Recomp, CompAct, EXIT). It further benefits from improved retrieval techniques such as query rewriting, underscoring that generation-side enhancements address limitations that retrieval alone cannot overcome.

Human mobility is crucial for urban planning (e.g., public transportation) and epidemic response strategies. However, existing research often neglects integrating comprehensive perspectives on spatial dynamics, temporal trends, and other contextual views due to the limitations of existing mobility datasets. To bridge this gap, we introduce MOBINS (MOBIlity Networked time Series), a novel dataset collection designed for networked time-series forecasting of dynamic human movements. MOBINS features diverse and explainable datasets that capture various mobility patterns across different transportation modes in four cities and two countries and cover both transportation and epidemic domains at the administrative area level. Our experiments with nine baseline methods reveal the significant impact of different model backbones on the proposed six datasets. We provide a valuable resource for advancing urban mobility research.

Understanding human mobility patterns is a complex challenge that requires modeling both node-oriented time series (e.g., population) and edge-oriented time series (e.g., population flows) within graph topologies across time.While previous methods have focused on either node-oriented time series or interactions, the synergistic integration of these two modalities has proven difficult to achieve.In this paper, we propose BINTS (BI-modal learning for Networked Time Series), a novel bi-modal learning framework that employs soft contrastive learning along the temporal axis.BINTS captures modality similarities and temporal patterns by simultaneously learning from evolving node-oriented time series and interactions, solving the limitations of single-modality approaches.To evaluate our method, we curate comprehensive multi-modal human mobility datasets spanning diverse locations and times.Our experimental results demonstrate that BINTS significantly outperforms existing forecasting models by capturing synergies across different data modalities.Overall, we establish BINTS as a powerful technique for holistically understanding and forecasting complex mobility dynamics.For reproducibility, the source code of our framework is available at https://github.com/kaist-dmlab/BINTS.

Abstract The application of language models in essay scoring has gained significant attention in recent years, typically on the basis of evaluating a single model across multiple prompts. However, in a multi-prompt setup, it is crucial to understand the varying aspects of different prompts. In such settings, there exist notable variations even in a trait with the same name across prompts. This semantic variation on the same traits underscores the need to treat them differently at a fine-grained level according to each prompt. In this study, we propose a multi-level disentanglement framework for multi-prompt essay scoring, designed to achieve fine-grained disentanglement of semantic differences across such traits. Our method not only improves the quality of the essay scoring, but also reduces memory usage and latency. Experimental results highlight that our framework surpasses seven state-of-the-art essay scoring methods and large language model(LLM)-based zero-shot and few-shot approaches, achieving the highest agreement with human essay ratings.

In light of the remarkable advancements made in time-series anomaly detection(TSAD), recent emphasis has been placed on exploiting the frequency domain as well as the time domain to address the difficulties in precisely detecting pattern-wise anomalies. However, in terms of anomaly scores, the window granularity of the frequency domain is inherently distinct from the data-point granularity of the time domain. Owing to this discrepancy, the anomaly information in the frequency domain has not been utilized to its full potential for TSAD. In this paper, we propose a TSAD framework, Dual-TF, that simultaneously uses both the time and frequency domains while breaking the time-frequency granularity discrepancy. To this end, our framework employs nested-sliding windows, with the outer and inner windows responsible for the time and frequency domains, respectively, and aligns the anomaly scores of the two domains. As a result of the high resolution of the aligned scores, the boundaries of pattern-wise anomalies can be identified more precisely. In six benchmark datasets, our framework outperforms state-of-the-art methods by 12.0--147%, as demonstrated by experimental results.

Robustness is becoming another important challenge of federated learning in that the data collection process in each client is naturally accompanied by noisy labels. However, it is far more complex and challenging owing to varying levels of data heterogeneity and noise over clients, which exacerbates the client-to-client performance discrepancy. In this work, we propose a robust federated learning method called FedRN, which exploits k-reliable neighbors with high data expertise or similarity. Our method helps mitigate the gap between low- and high-performance clients by training only with a selected set of clean examples, identified by a collaborative model that is built based on the reliability score over clients. We demonstrate the superiority of FedRN via extensive evaluations on three real-world or synthetic benchmark datasets. Compared with existing robust methods, the results show that FedRN significantly improves the test accuracy in the presence of noisy labels.

Cross-domain few-shot learning (CD-FSL), where there are few target samples under extreme differences between source and target domains, has recently attracted huge attention. Recent studies on CD-FSL generally focus on transfer learning based approaches, where a neural network is pre-trained on popular labeled source domain datasets and then transferred to target domain data. Although the labeled datasets may provide suitable initial parameters for the target data, the domain difference between the source and target might hinder fine-tuning on the target domain. This paper proposes a simple yet powerful method that re-randomizes the parameters fitted on the source domain before adapting to the target data. The re-randomization resets source-specific parameters of the source pre-trained model and thus facilitates fine-tuning on the target domain, improving few-shot performance.

Understanding vision and language representations of product content is vital for search and recommendation applications in e-commerce. As a backbone for online shopping platforms and inspired by the recent success in representation learning research, we propose a contrastive learning framework that aligns language and visual models using unlabeled raw product text and images. We present techniques we used to train large-scale representation learning models and share solutions that address domain-specific challenges. We study the performance using our pre-trained model as backbones for diverse downstream tasks, including category classification, attribute extraction, product matching, product clustering, and adult product recognition. Experimental results show that our proposed method outperforms the baseline in each downstream task regarding both single modality and multiple modalities.

Digital contact tracing is an effective solution to prevent such a pandemic, but the low adoption rate of a required mobile app hinders its effectiveness. A large collection of cellular trajectories from mobile subscribers can be an out-of-the-box solution that is free from the low adoption issue, but has been overlooked due to its low spatial resolution. In this paper, to increase the resolution of this cellular trajectory, we present a new problem that estimates the user’s visited places at the point-of-interest(POI) level, which we call POI-level cellular trajectory reconstruction. We propose a novel algorithm, Pincette, that accomplishes more accurate POI reconstruction by leveraging various external data such as road networks and POI contexts. Specifically, Pincette comprises multi-view feature extraction and GCN-LSTM-based POI estimation. In the multi-view feature extraction, Pincette extracts three complementary features from three views: efficiency, periodicity, and popularity. In the GCN-LSTM-based POI estimation, these three views are seamlessly integrated, where spatio-temporal periodic patterns are captured by graph convolutional networks (GCNs) and an LSTM. With extensive experiments on two real data collections of two cities, we show that Pincette outperforms four POI estimation baselines by up to 21.20%. We believe that our work sheds light on the use of cellular trajectories for digital contact tracing. We release the source code at https://github.com/kaist-dmlab/Pincette.

Responsible AI becomes critical where robustness and fairness must be satisfied together. Traditionally, the two topics have been studied by different communities for different applications. Robust training is designed for noisy or poisoned data where image data is typically considered. In comparison, fair training primarily deals with biased data where structured data is typically considered. Nevertheless, robust training and fair training are fundamentally similar in considering that both of them aim at fixing the inherent flaws of real-world data. In this tutorial, we first cover state-of-the-art robust training techniques where most of the research is on combating various label noises. In particular, we cover label noise modeling, robust training approaches, and real-world noisy data sets. Then, proceeding to the related fairness literature, we discuss pre-processing, in-processing, and post-processing unfairness mitigation techniques, depending on whether the mitigation occurs before, during, or after the model training. Finally, we cover the recent trend emerged to combine robust and fair training in two flavors: the former is to make the fair training more robust (i.e., robust fair training), and the latter is to consider robustness and fairness as two equals to incorporate them into a holistic framework. This tutorial is indeed timely and novel because the convergence of the two topics is increasingly common, but yet to be addressed in tutorials. The tutors have extensive experience publishing papers in top-tier machine learning and data mining venues and developing machine learning platforms.

Real-world data inevitably contains noisy labels, which induce the poor generalization of deep neural networks. It is known that the network typically begins to rapidly memorize false-labeled samples after a certain point of training. Thus, to counter the label noise challenge, we propose a novel self-transitional learning method called MORPH, which automatically switches its learning phase at the transition point from seeding to evolution. In the seeding phase, the network is updated using all the samples to collect a seed of clean samples. Then, in the evolution phase, the network is updated using only the set of arguably clean samples, which precisely keeps expanding by the updated network. Thus, MORPH effectively avoids the overfitting to false-labeled samples throughout the entire training period. Extensive experiments using five real-world or synthetic benchmark datasets demonstrate substantial improvements over state-of-the-art methods in terms of robustness and efficiency.

The accuracy of deep neural networks is significantly affected by how well mini-batches are constructed during the training step. In this paper, we propose a novel adaptive batch selection algorithm called Recency Bias that exploits the uncertain samples predicted inconsistently in recent iterations. The historical label predictions of each training sample are used to evaluate its predictive uncertainty within a sliding window. Then, the sampling probability for the next mini-batch is assigned to each training sample in proportion to its predictive uncertainty. By taking advantage of this design, Recency Bias not only accelerates the training step but also achieves a more accurate network. We demonstrate the superiority of Recency Bias by extensive evaluation on two independent tasks. Compared with existing batch selection methods, the results showed that Recency Bias reduced the test error by up to 20.97% in a fixed wall-clock training time. At the same time, it improved the training time by up to 59.32% to reach the same test error.

The escalating crisis of COVID-19 has put people all over the world in danger. Owing to the high contagion rate of the virus, COVID-19 cases continue to increase globally. To further suppress the threat of the COVID-19 pandemic and minimize its damage, it is imperative that each country monitors inbound travelers. Moreover, given that resources for quarantine are often limited, they must be carefully allocated. In this paper, to aid in such allocation by predicting the number of inbound COVID-19 cases, we propose Hi-COVIDNet, which takes advantage of the geographic hierarchy. Hi-COVIDNet is based on a neural network with two-level components, namely, country-level and continent-level encoders, which understand the complex relationships among foreign countries and derive their respective contagion risk to the destination country. An in-depth case study in South Korea with real-world COVID-19 datasets confirmed the effectiveness and practicality of Hi-COVIDNet.

Recently, autoencoder (AE)-based embedding approaches have achieved state-of-the-art performance in many tasks, especially in top-k recommendation with user embedding or node classification with node embedding. However, we find that many real-world data follow the power-law distribution with respect to the data object sparsity. When learning AE-based embeddings of these data, dense inputs move away from sparse inputs in an embedding space even when they are highly correlated. This phenomenon, which we call polarization, obviously distorts the embedding. In this paper, we propose TRAP that leverages two-level regularizers to effectively alleviate the polarization problem. The macroscopic regularizer generally prevents dense input objects from being distant from other sparse input objects, and the microscopic regularizer individually attracts each object to correlated neighbor objects rather than uncorrelated ones. Importantly, TRAP is a meta-algorithm that can be easily coupled with existing AE-based embedding methods with a simple modification. In extensive experiments on two representative embedding tasks using six-real world datasets, TRAP boosted the performance of the state-of-the-art algorithms by up to 31.53% and 94.99% respectively.

In most parallel DBSCAN algorithms, neighboring points are assigned to the same data partition for parallel processing to facilitate calculation of the density of the neighbors. This data partitioning scheme causes a few critical problems including load imbalance between data partitions, especially in a skewed data set. To remedy these problems, we propose a cell-based data partitioning scheme, pseudo random partitioning , that randomly distributes small cells rather than the points themselves. It achieves high load balance regardless of data skewness while retaining the data contiguity required for DBSCAN. In addition, we build and broadcast a highly compact summary of the entire data set, which we call a two-level cell dictionary , to supplement random partitions. Then, we develop a novel parallel DBSCAN algorithm, Random Partitioning-DBSCAN (shortly, RP-DBSCAN), that uses pseudo random partitioning together with a two-level cell dictionary. The algorithm simultaneously finds the local clusters to each data partition and then merges these local clusters to obtain global clustering. To validate the merit of our approach, we implement RP-DBSCAN on Spark and conduct extensive experiments using various real-world data sets on 12 Microsoft Azure machines (48 cores). In RP-DBSCAN, data partitioning and cluster merging are very light, and clustering on each split is not dragged out by a specific worker. Therefore, the performance results show that RP-DBSCAN significantly outperforms the state-of-the-art algorithms by up to 180 times.

The k-medoids algorithm is one of the best-known clustering algorithms. Despite this, however, it is not as widely used for big data analytics as the k-means algorithm, mainly because of its high computational complexity. Many studies have attempted to solve the efficiency problem of the k-medoids algorithm, but all such studies have improved efficiency at the expense of accuracy. In this paper, we propose a novel parallel k-medoids algorithm, which we call PAMAE, that achieves both high accuracy and high efficiency. We identify two factors---"global search" and "entire data"---that are essential to achieving high accuracy, but are also very time-consuming if considered simultaneously. Thus, our key idea is to apply them individually through two phases: parallel seeding and parallel refinement, neither of which is costly. The first phase performs global search over sampled data, and the second phase performs local search over entire data. Our theoretical analysis proves that this serial execution of the two phases leads to an accurate solution that would be achieved by global search over entire data. In order to validate the merit of our approach, we implement PAMAE on Spark as well as Hadoop and conduct extensive experiments using various real-world data sets on 12 Microsoft Azure machines (48 cores). The results show that PAMAE significantly outperforms most of recent parallel algorithms and, at the same time, produces a clustering quality as comparable as the previous most-accurate algorithm. The source code and data are available at https://github.com/jaegil/k-Medoid.