GUI agents powered by LLMs show promise in interacting with diverse digital environments.Among these, video games offer a valuable testbed due to their varied interfaces, with adventure games posing additional challenges through complex, narrative-driven interactions.Existing game benchmarks, however, lack diversity and rarely evaluate agents on completing entire storylines.To address this, we introduce FlashAdventure, a benchmark of 34 Flashbased adventure games designed to test full story arc completion and tackle the observationbehavior gap: the challenge of remembering and acting on earlier gameplay information.We also propose CUA-as-a-Judge, an automated gameplay evaluator, and COAST, an agentic framework leveraging long-term clue memory to better plan and solve sequential tasks.Experiments show current GUI agents struggle with full story arcs, while COAST improves milestone completion by bridging the observationbehavior gap.Nonetheless, a marked discrepancy between humans and best-performing agents warrants continued research efforts to narrow this divide. * Equal contribution. †Work done during an internship at KRAFTON. Flash-Based Adventure GamesInput GUI Agent (Operator) Gameplay

Diffusion models excel in generative tasks, but aligning them with specific objectives while maintaining their versatility remains challenging. Existing fine-tuning methods often suffer from reward over-optimization, while approximate guidance approaches fail to optimize target rewards effectively. Addressing these limitations, we propose a training-free, test-time method based on Sequential Monte Carlo (SMC) to sample from the reward-aligned target distribution. Our approach, tailored for diffusion sampling and incorporating tempering techniques, achieves comparable or superior target rewards to fine-tuning methods while preserving diversity and cross-reward generalization. We demonstrate its effectiveness in single-reward optimization, multi-objective scenarios, and online black-box optimization. This work offers a robust solution for aligning diffusion models with diverse downstream objectives without compromising their general capabilities. Code is available at https://github.com/krafton-ai/DAS.

State-of-the-art text-to-image (T2I) diffusion models often struggle to generate rare compositions of concepts, e.g., objects with unusual attributes. In this paper, we show that the compositional generation power of diffusion models on such rare concepts can be significantly enhanced by the Large Language Model (LLM) guidance. We start with empirical and theoretical analysis, demonstrating that exposing frequent concepts relevant to the target rare concepts during the diffusion sampling process yields more accurate concept composition. Based on this, we propose a training-free approach, R2F, that plans and executes the overall rare-to-frequent concept guidance throughout the diffusion inference by leveraging the abundant semantic knowledge in LLMs. Our framework is flexible across any pre-trained diffusion models and LLMs, and can be seamlessly integrated with the region-guided diffusion approaches. Extensive experiments on three datasets, including our newly proposed benchmark, RareBench, containing various prompts with rare compositions of concepts, R2F significantly surpasses existing models including SD3.0 and FLUX by up to 28.1%p in T2I alignment. Code is available at https://github.com/krafton-ai/Rare-to-Frequent.

*Continual learning (CL)* enables deep neural networks to adapt to ever-changing data distributions. In practice, there may be scenarios where annotation is costly, leading to *active continual learning (ACL)*, which performs *active learning (AL)* for the CL scenarios when reducing the labeling cost by selecting the most informative subset is preferable. However, conventional AL strategies are not suitable for ACL, as they focus solely on learning the new knowledge, leading to *catastrophic forgetting* of previously learned tasks. Therefore, ACL requires a new AL strategy that can balance the prevention of catastrophic forgetting and the ability to quickly learn new tasks. In this paper, we propose **AccuACL**, **Accu**mulated informativeness-based **A**ctive **C**ontinual **L**earning, by the novel use of the Fisher information matrix as a criterion for sample selection, derived from a theoretical analysis of the Fisher-optimality preservation properties within the framework of ACL, while also addressing the scalability issue of Fisher information-based AL. Extensive experiments demonstrate that AccuACL significantly outperforms AL baselines across various CL algorithms, increasing the average accuracy and forgetting by 23.8% and 17.0%, respectively, on average.

We propose a novel framework DropTop that suppresses the shortcut bias in online continual learning (OCL) while being adaptive to the varying degree of the shortcut bias incurred by continuously changing environment. By the observed high-attention property of the shortcut bias, highly-activated features are considered candidates for debiasing. More importantly, resolving the limitation of the online environment where prior knowledge and auxiliary data are not ready, two novel techniques---feature map fusion and adaptive intensity shifting---enable us to automatically determine the appropriate level and proportion of the candidate shortcut features to be dropped. Extensive experiments on five benchmark datasets demonstrate that, when combined with various OCL algorithms, DropTop increases the average accuracy by up to 10.4% and decreases the forgetting by up to 63.2%.

In real-world continual learning (CL) scenarios, tasks often exhibit intricate and unpredictable semantic shifts, posing challenges for fixed prompt management strategies which are tailored to only handle semantic shifts of uniform degree (i.e., uniformly mild or uniformly abrupt). To address this limitation, we propose an adaptive prompting approach that effectively accommodates semantic shifts of varying degree where mild and abrupt shifts are mixed. AdaPromptCL employs the assign-and-refine semantic grouping mechanism that dynamically manages prompt groups in accordance with the semantic similarity between tasks, enhancing the quality of grouping through continuous refinement. Our experiment results demonstrate that AdaPromptCL outperforms existing prompting methods by up to 21.3%, especially in the benchmark datasets with diverse semantic shifts between tasks.

Labels are typically sparse in real-world time series due to the high annotation cost. Recently, consistency regularization techniques have been used to generate artificial labels from unlabeled augmented instances. To fully exploit the sequential characteristic of time series in consistency regularization, we propose a novel method of data augmentation called *context-attached augmentation*, which adds preceding and succeeding instances to a target instance to form its augmented instance. Unlike the existing augmentation techniques that modify a target instance by directly perturbing its attributes, the context-attached augmentation generates instances augmented with varying contexts while maintaining the target instance. Based on our augmentation method, we propose a *context consistency regularization* framework, which first adds different contexts to a target instance sampled from a given time series and then shares unitary reliability-based cross-window labels across the augmented instances to maintain consistency. We demonstrate that the proposed framework outperforms the existing state-of-the-art consistency regularization frameworks through comprehensive experiments on real-world time-series datasets.

Data pruning, which aims to downsize a large training set into a small informative subset, is crucial for reducing the enormous computational costs of modern deep learning. Though large-scale data collections invariably contain annotation noise and numerous robust learning methods have been developed, data pruning for the noise-robust learning scenario has received little attention. With state-of-the-art Re-labeling methods that self-correct erroneous labels while training, it is challenging to identify which subset induces the most accurate re-labeling of erroneous labels in the entire training set. In this paper, we formalize the problem of data pruning with re-labeling. We first show that the likelihood of a training example being correctly re-labeled is proportional to the prediction confidence of its neighborhood in the subset. Therefore, we propose a novel data pruning algorithm, Prune4Rel, that finds a subset maximizing the total neighborhood confidence of all training examples, thereby maximizing the re-labeling accuracy and generalization performance. Extensive experiments on four real and one synthetic noisy datasets show that Prune4Rel outperforms the baselines with Re-labeling models by up to 9.1% as well as those with a standard model by up to 21.6%.

Deep learning has achieved remarkable success in numerous domains with help from large amounts of big data. However, the quality of data labels is a concern because of the lack of high-quality labels in many real-world scenarios. As noisy labels severely degrade the generalization performance of deep neural networks, learning from noisy labels (robust training) is becoming an important task in modern deep learning applications. In this survey, we first describe the problem of learning with label noise from a supervised learning perspective. Next, we provide a comprehensive review of 62 state-of-the-art robust training methods, all of which are categorized into five groups according to their methodological difference, followed by a systematic comparison of six properties used to evaluate their superiority. Subsequently, we perform an in-depth analysis of noise rate estimation and summarize the typically used evaluation methodology, including public noisy datasets and evaluation metrics. Finally, we present several promising research directions that can serve as a guideline for future studies.

Online recommender systems should be always aligned with users' current interest to accurately suggest items that each user would like. Since user interest usually evolves over time, the update strategy should be flexible to quickly catch users' current interest from continuously generated new user-item interactions. Existing update strategies focus either on the importance of each user-item interaction or the learning rate for each recommender parameter, but such one-directional flexibility is insufficient to adapt to varying relationships between interactions and parameters. In this paper, we propose MeLON, a meta-learning based novel online recommender update strategy that supports two-directional flexibility. It is featured with an adaptive learning rate for each parameter-interaction pair for inducing a recommender to quickly learn users' up-to-date interest. The procedure of MeLON is optimized following a meta-learning approach: it learns how a recommender learns to generate the optimal learning rates for future updates. Specifically, MeLON first enriches the meaning of each interaction based on previous interactions and identifies the role of each parameter for the interaction; and then combines these two pieces of information to generate an adaptive learning rate. Theoretical analysis and extensive evaluation on three real-world online recommender datasets validate the effectiveness of MeLON.

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.

Unlabeled data examples awaiting annotations contain open-set noise inevitably. A few active learning studies have attempted to deal with this open-set noise for sample selection by filtering out the noisy examples. However, because focusing on the purity of examples in a query set leads to overlooking the informativeness of the examples, the best balancing of purity and informativeness remains an important question. In this paper, to solve this purity-informativeness dilemma in open-set active learning, we propose a novel Meta-Query-Net (MQ-Net) that adaptively finds the best balancing between the two factors. Specifically, by leveraging the multi-round property of active learning, we train MQ-Net using a query set without an additional validation set. Furthermore, a clear dominance relationship between unlabeled examples is effectively captured by MQ-Net through a novel skyline regularization. Extensive experiments on multiple open-set active learning scenarios demonstrate that the proposed MQ-Net achieves 20.14% improvement in terms of accuracy, compared with the state-of-the-art methods.

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.

A deep neural network (DNN) has achieved great success in many machine learning tasks by virtue of its high expressive power. However, its prediction can be easily biased to undesirable features, which are not essential for solving the target task and are even imperceptible to a human, thereby resulting in poor generalization. Leveraging plenty of undesirable features in out-of-distribution (OOD) examples has emerged as a potential solution for de-biasing such features, and a recent study shows that softmax-level calibration of OOD examples can successfully remove the contribution of undesirable features to the last fully-connected layer of a classifier. However, its applicability is confined to the classification task, and its impact on a DNN feature extractor is not properly investigated. In this paper, we propose Taufe, a novel regularizer that deactivates many undesirable features using OOD examples in the feature extraction layer and thus removes the dependency on the task-specific softmax layer. To show the task-agnostic nature of Taufe, we rigorously validate its performance on three tasks, classification, regression, and a mix of them, on CIFAR-10, CIFAR-100, ImageNet, CUB200, and CAR datasets. The results demonstrate that Taufe consistently outperforms the state-of-the-art method as well as the baselines without regularization.

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.

Catastrophic forgetting is a critical challenge in training deep neural networks. Although continual learning has been investigated as a countermeasure to the problem, it often suffers from the requirements of additional network components and the limited scalability to a large number of tasks. We propose a novel approach to continual learning by approximating a true loss function using an asymmetric quadratic function with one of its sides overestimated. Our algorithm is motivated by the empirical observation that the network parameter updates affect the target loss functions asymmetrically. In the proposed continual learning framework, we estimate an asymmetric loss function for the tasks considered in the past through a proper overestimation of its unobserved sides in training new tasks, while deriving the accurate model parameter for the observable sides. In contrast to existing approaches, our method is free from the side effects and achieves the state-of-the-art accuracy that is even close to the upper-bound performance on several challenging benchmark datasets.

A power-optimized low-voltage VCO with wide tuning range is proposed. The VCO achieves low power by optimum selection of inductance in the L-C tank. Despite the low power supply near threshold voltage, the VCO achieves wide tuning range by using a voltage-boosted digital tuning technique. It is shown that the power consumption of VCO with a required output swing and phase noise can be minimized if the inductance is chosen such that it's LQ and Q/L is maximized. To increase the tuning range, a digital tuning method is used where the output voltage of the VCO is exploited to generate a high voltage for the switches in the digital tuning scheme. The proposed VCO achieves phase noise of -120 dBc/Hz at 1-MHz offset and 18 % tuning range while consuming 660 /spl mu/A in 0.5-V supply. Figure-of-merit with tuning range of the proposed VCO is -197.1 dB, which is the lowest among the recent state-of-the-art low-voltage VCOs.