Sungeun Hong

Weizheng Wang, Obi Ike, Soyun Choi et al.

Social robot navigation increasingly relies on large language models for reasoning, path planning, and enabling movement in dynamic human spaces. However, relying solely on LLMs for planning often leads to unpredictable and unsafe behaviors, especially in dynamic human spaces, due to limited physical grounding and weak logical consistency. In this work, we introduce NaviWM, a socially-aware robot Navigation World Model that augments LLM reasoning with a structured world model and a logic-driven chain-of-thought process. NaviWM consists of two main components: (1) a spatial-temporal world model that captures the positions, velocities, and activities of agents in the environment, and (2) a deductive reasoning module that guides LLMs through a multi-step, logic-based inference process. This integration enables the robot to generate navigation decisions that are both socially compliant and physically safe, under well-defined constraints such as personal space, collision avoidance, and timing. Unlike previous methods based on prompting or fine-tuning, NaviWM encodes social norms as first-order logic, enabling interpretable and verifiable reasoning. Experiments show that NaviWM improves success rates and reduces social violations, particularly in crowded environments. These results demonstrate the benefit of combining formal reasoning with LLMs for robust social navigation. Additional experimental details and demo videos for this work can be found at: https://sites.google.com/view/NaviWM.

Youjia Zhang, Soyun Choi, Sungeun Hong

Crowd counting research has made significant advancements in real-world applications, but it remains a formidable challenge in cross-modal settings. Most existing methods rely solely on the optical features of RGB images, ignoring the feasibility of other modalities such as thermal and depth images. The inherently significant differences between the different modalities and the diversity of design choices for model architectures make cross-modal crowd counting more challenging. In this paper, we propose Cross-modal Spatio-Channel Attention (CSCA) blocks, which can be easily integrated into any modality-specific architecture. The CSCA blocks first spatially capture global functional correlations among multi-modality with less overhead through spatial-wise cross-modal attention. Cross-modal features with spatial attention are subsequently refined through adaptive channel-wise feature aggregation. In our experiments, the proposed block consistently shows significant performance improvement across various backbone networks, resulting in state-of-the-art results in RGB-T and RGB-D crowd counting.

Byungho Jo, Donghyeon Cho, In Kyu Park et al.

Existing face restoration models have relied on general assessment metrics that do not consider the characteristics of facial regions. Recent works have therefore assessed their methods using human studies, which is not scalable and involves significant effort. This paper proposes a novel face-centric metric based on an adversarial framework where a generator simulates face restoration and a discriminator assesses image quality. Specifically, our per-pixel discriminator enables interpretable evaluation that cannot be provided by traditional metrics. Moreover, our metric emphasizes facial primary regions considering that even minor changes to the eyes, nose, and mouth significantly affect human cognition. Our face-oriented metric consistently surpasses existing general or facial image quality assessment metrics by impressive margins. We demonstrate the generalizability of the proposed strategy in various architectural designs and challenging scenarios. Interestingly, we find that our IFQA can lead to performance improvement as an objective function.

Kyeongtak Han, Youngeun Kim, Dongyoon Han et al.

Active domain adaptation (ADA) studies have mainly addressed query selection while following existing domain adaptation strategies. However, we argue that it is critical to consider not only query selection criteria but also domain adaptation strategies designed for ADA scenarios. This paper introduces sequential learning considering both domain type (source/target) or labelness (labeled/unlabeled). We first train our model only on labeled target samples obtained by loss-based query selection. When loss-based query selection is applied under domain shift, unuseful high-loss samples gradually increase, and the labeled-sample diversity becomes low. To solve these, we fully utilize pseudo labels of the unlabeled target domain by leveraging loss prediction. We further encourage pseudo labels to have low self-entropy and diverse class distributions. Our model significantly outperforms previous methods as well as baseline models in various benchmark datasets.

Hongyeob Kim, Inyoung Jung, Dayoon Suh et al.

Audio-Visual Question Answering (AVQA) requires not only question-based multimodal reasoning but also precise temporal grounding to capture subtle dynamics for accurate prediction. However, existing methods mainly use question information implicitly, limiting focus on question-specific details. Furthermore, most studies rely on uniform frame sampling, which can miss key question-relevant frames. Although recent Top-K frame selection methods aim to address this, their discrete nature still overlooks fine-grained temporal details. This paper proposes QA-TIGER, a novel framework that explicitly incorporates question information and models continuous temporal dynamics. Our key idea is to use Gaussian-based modeling to adaptively focus on both consecutive and non-consecutive frames based on the question, while explicitly injecting question information and applying progressive refinement. We leverage a Mixture of Experts (MoE) to flexibly implement multiple Gaussian models, activating temporal experts specifically tailored to the question. Extensive experiments on multiple AVQA benchmarks show that QA-TIGER consistently achieves state-of-the-art performance. Code is available at https://aim-skku.github.io/QA-TIGER/

Aecheon Jung, Seunghwan Lee, Dongyoon Han et al.

Model merging offers an efficient alternative to multi-task learning by combining independently fine-tuned models, but most prior approaches focus mainly on avoiding task interference. We argue instead that the real potential of merging lies in achieving synergy, where tasks enhance one another. Our intuition comes from a pilot study showing that when a classifier trained on one task is paired with the encoder of another, the resulting cross-task performance strongly predicts merge quality. Moreover, adapting even a single task-specific layer can substantially improve this compatibility, suggesting a simple yet powerful lever for synergy. Building on this insight, we introduce SyMerge, a lightweight framework that jointly optimizes one task-specific layer and merging coefficients. To ensure stability without labels, SyMerge employs a robust self-labeling strategy guided by expert model predictions, avoiding the pitfalls of entropy-based adaptation. This minimalist yet principled design achieves state-of-the-art results across vision, dense prediction, and NLP benchmarks, while also producing adapted layers that transfer effectively to other merging methods. Our code is available at https://aim-skku.github.io/SyMerge/

Yoorhim Cho, Hongyeob Kim, Semin Kim et al.

Visuo-tactile perception aims to understand an object's tactile properties, such as texture, softness, and rigidity. However, the field remains underexplored because collecting tactile data is costly and labor-intensive. We observe that visually distinct objects can exhibit similar surface textures or material properties. For example, a leather sofa and a leather jacket have different appearances but share similar tactile properties. This implies that tactile understanding can be guided by material cues in visual data, even without direct tactile supervision. In this paper, we introduce RA-Touch, a retrieval-augmented framework that improves visuo-tactile perception by leveraging visual data enriched with tactile semantics. We carefully recaption a large-scale visual dataset with tactile-focused descriptions, enabling the model to access tactile semantics typically absent from conventional visual datasets. A key challenge remains in effectively utilizing these tactile-aware external descriptions. RA-Touch addresses this by retrieving visual-textual representations aligned with tactile inputs and integrating them to focus on relevant textural and material properties. By outperforming prior methods on the TVL benchmark, our method demonstrates the potential of retrieval-based visual reuse for tactile understanding. Code is available at https://aim-skku.github.io/RA-Touch

Youngeun Kim, Donghyeon Cho, Kyeongtak Han et al.

Domain adaptation assumes that samples from source and target domains are freely accessible during a training phase. However, such an assumption is rarely plausible in the real-world and possibly causes data-privacy issues, especially when the label of the source domain can be a sensitive attribute as an identifier. To avoid accessing source data that may contain sensitive information, we introduce Source data-Free Domain Adaptation (SFDA). Our key idea is to leverage a pre-trained model from the source domain and progressively update the target model in a self-learning manner. We observe that target samples with lower self-entropy measured by the pre-trained source model are more likely to be classified correctly. From this, we select the reliable samples with the self-entropy criterion and define these as class prototypes. We then assign pseudo labels for every target sample based on the similarity score with class prototypes. Furthermore, to reduce the uncertainty from the pseudo labeling process, we propose set-to-set distance-based filtering which does not require any tunable hyperparameters. Finally, we train the target model with the filtered pseudo labels with regularization from the pre-trained source model. Surprisingly, without direct usage of labeled source samples, our PrDA outperforms conventional domain adaptation methods on benchmark datasets. Our code is publicly available at https://github.com/youngryan1993/SFDA-SourceFreeDA

Sungeun Hong, Woobin Im, Hyun S. Yang

Up to now, only limited research has been conducted on cross-modal retrieval of suitable music for a specified video or vice versa. Moreover, much of the existing research relies on metadata such as keywords, tags, or associated description that must be individually produced and attached posterior. This paper introduces a new content-based, cross-modal retrieval method for video and music that is implemented through deep neural networks. We train the network via inter-modal ranking loss such that videos and music with similar semantics end up close together in the embedding space. However, if only the inter-modal ranking constraint is used for embedding, modality-specific characteristics can be lost. To address this problem, we propose a novel soft intra-modal structure loss that leverages the relative distance relationship between intra-modal samples before embedding. We also introduce reasonable quantitative and qualitative experimental protocols to solve the lack of standard protocols for less-mature video-music related tasks. Finally, we construct a large-scale 200K video-music pair benchmark. All the datasets and source code can be found in our online repository (https://github.com/csehong/VM-NET).

Sungeun Hong, Woobin Im, Jongbin Ryu et al.

Real-world face recognition using a single sample per person (SSPP) is a challenging task. The problem is exacerbated if the conditions under which the gallery image and the probe set are captured are completely different. To address these issues from the perspective of domain adaptation, we introduce an SSPP domain adaptation network (SSPP-DAN). In the proposed approach, domain adaptation, feature extraction, and classification are performed jointly using a deep architecture with domain-adversarial training. However, the SSPP characteristic of one training sample per class is insufficient to train the deep architecture. To overcome this shortage, we generate synthetic images with varying poses using a 3D face model. Experimental evaluations using a realistic SSPP dataset show that deep domain adaptation and image synthesis complement each other and dramatically improve accuracy. Experiments on a benchmark dataset using the proposed approach show state-of-the-art performance. All the dataset and the source code can be found in our online repository (https://github.com/csehong/SSPP-DAN).

Seunghwan Lee, Inyoung Jung, Hojoon Lee et al.

Continual Test-Time Adaptation (CTTA) enables pre-trained models to adapt to continuously evolving domains. Existing methods have improved robustness but typically rely on fixed or batch-level thresholds, which cannot account for varying difficulty across classes and instances. This limitation is especially problematic in semantic segmentation, where each image requires dense, multi-class predictions. We propose an approach that adaptively adjusts pseudo labels to reflect the confidence distribution within each image and dynamically balances learning toward classes most affected by domain shifts. This fine-grained, class- and instance-aware adaptation produces more reliable supervision and mitigates error accumulation throughout continual adaptation. Extensive experiments across eight CTTA and TTA scenarios, including synthetic-to-real and long-term shifts, show that our method consistently outperforms state-of-the-art techniques, setting a new standard for semantic segmentation under evolving conditions.

Jungbin Cho, Junwan Kim, Jisoo Kim et al.

Human motion is inherently continuous and dynamic, posing significant challenges for generative models. While discrete generation methods are widely used, they suffer from limited expressiveness and frame-wise noise artifacts. In contrast, continuous approaches produce smoother, more natural motion but often struggle to adhere to conditioning signals due to high-dimensional complexity and limited training data. To resolve this 'discord' between discrete and continuous representations we introduce DisCoRD: Discrete Tokens to Continuous Motion via Rectified Flow Decoding, a novel method that leverages rectified flow to decode discrete motion tokens in the continuous, raw motion space. Our core idea is to frame token decoding as a conditional generation task, ensuring that DisCoRD captures fine-grained dynamics and achieves smoother, more natural motions. Compatible with any discrete-based framework, our method enhances naturalness without compromising faithfulness to the conditioning signals on diverse settings. Extensive evaluations demonstrate that DisCoRD achieves state-of-the-art performance, with FID of 0.032 on HumanML3D and 0.169 on KIT-ML. These results establish DisCoRD as a robust solution for bridging the divide between discrete efficiency and continuous realism. Project website: https://whwjdqls.github.io/discord-motion/

Youngeun Kim, Seunghwan Lee, Aecheon Jung et al.

Model merging enables efficient multi-task models by combining task-specific fine-tuned checkpoints. However, storing multiple task-specific checkpoints requires significant memory, limiting scalability and restricting model merging to larger models and diverse tasks. In this paper, we propose quantizing task vectors (i.e., the difference between pre-trained and fine-tuned checkpoints) instead of quantizing fine-tuned checkpoints. We observe that task vectors exhibit a narrow weight range, enabling low precision quantization (e.g., 4 bit) within existing task vector merging frameworks. To further mitigate quantization errors within ultra-low bit precision (e.g., 2 bit), we introduce Residual Task Vector Quantization, which decomposes the task vector into a base vector and offset component. We allocate bits based on quantization sensitivity, ensuring precision while minimizing error within a memory budget. Experiments on image classification and dense prediction show our method maintains or improves model merging performance while using only 8% of the memory required for full-precision checkpoints.

Jinsu Shin, Sungeun Hong, Jin Yeong Bak

Online handwriting represents strokes as time-ordered trajectories, which makes handwritten content easier to transform and reuse in a wide range of applications. However, generating natural sentence-level online handwriting that faithfully reflects a writer's style remains challenging, since sentence synthesis demands context-dependent characters with stroke continuity and spacing. Prior methods treat these boundary properties as implicit outcomes of sequence modeling, which becomes unreliable at the sentence scale and under limited compositional diversity. We propose CASHG, a context-aware stylized online handwriting generator that explicitly models inter-character connectivity for style-consistent sentence-level trajectory synthesis. CASHG uses a Character Context Encoder to obtain character identity and sentence-dependent context memory and fuses them in a bigram-aware sliding-window Transformer decoder that emphasizes local predecessor--current transitions, complemented by gated context fusion for sentence-level context.Training proceeds through a three-stage curriculum from isolated glyphs to full sentences, improving robustness under sparse transition coverage. We further introduce Connectivity and Spacing Metrics (CSM), a boundary-aware evaluation suite that quantifies cursive connectivity and spacing similarity. Under benchmark-matched evaluation protocols, CASHG consistently improves CSM over comparison methods while remaining competitive in DTW-based trajectory similarity, with gains corroborated by a human evaluation.

Youngeun Kim, Youjia Zhang, Huiling Liu et al.

Large Vision-Language Models (VLMs) enable strong multimodal reasoning but incur heavy inference costs from redundant visual tokens. Token pruning alleviates this issue, yet existing approaches face limitations. Attention-based methods rely on raw attention scores, which are often unstable across layers and heads and can lead to redundant selections. Diversity-based methods improve robustness by selecting tokens far apart in feature space but risk dropping regions needed for accurate prediction. We propose \ours, a training-free framework built on a simple intuition: tokens with higher sensitivity are more likely to influence the model's output, and they should also capture complementary visual cues rather than overlapping information. To achieve this, we estimate token sensitivity using zeroth-order perturbations at the projection layer, a shallow and computationally light component of the model. This approach measures how small random perturbations affect the projection outputs, allowing us to approximate each token's influence through lightweight forward passes without backpropagation. Extensive experiments across multiple VLMs and benchmarks show that \ours consistently outperforms prior methods, pruning up to 94.4\% of tokens while maintaining accuracy and significantly improving efficiency, achieving up to 2.30x faster end-to-end inference over the baseline.

Youjia Zhang, Youngeun Kim, Young-Geun Choi et al.

Test-time adaptation (TTA) enhances the zero-shot robustness under distribution shifts by leveraging unlabeled test data during inference. Despite notable advances, several challenges still limit its broader applicability. First, most methods rely on backpropagation or iterative optimization, which limits scalability and hinders real-time deployment. Second, they lack explicit modeling of class-conditional feature distributions. This modeling is crucial for producing reliable decision boundaries and calibrated predictions, but it remains underexplored due to the lack of both source data and supervision at test time. In this paper, we propose ADAPT, an Advanced Distribution-Aware and backPropagation-free Test-time adaptation method. We reframe TTA as a Gaussian probabilistic inference task by modeling class-conditional likelihoods using gradually updated class means and a shared covariance matrix. This enables closed-form, training-free inference. To correct potential likelihood bias, we introduce lightweight regularization guided by CLIP priors and a historical knowledge bank. ADAPT requires no source data, no gradient updates, and no full access to target data, supporting both online and transductive settings. Extensive experiments across diverse benchmarks demonstrate that our method achieves state-of-the-art performance under a wide range of distribution shifts with superior scalability and robustness.

Hyuntak Shin, Aecheon Jung, Sungeun Hong et al.

Low-rank training methods reduce the number of trainable parameters by re-parameterizing the weights with matrix decompositions (e.g., singular value decomposition). However, enforcing a fixed low-rank structure caps the rank of the weight matrices and can hinder the model's ability to learn complex patterns. Furthermore, the effective rank of the model's weights tends to decline during training, and this drop is accelerated when the model is reparameterized into a low-rank structure. In this study, we argue that strategically interleaving full-rank training epochs within low-rank training epochs can effectively restore the rank of the model's weights. Based on our findings, we propose a general dynamic-rank training framework that is readily applicable to a wide range of neural-network tasks. We first describe how to adjust the rank of weight matrix to alleviate the inevitable rank collapse that arises during training, and then present extensive empirical results that validate our claims and demonstrate the efficacy of the proposed framework. Our empirical study shows that the proposed method achieves almost the same computational cost as SVD-based low-rank training while achieving a comparable accuracy to full-rank training across various benchmarks.

Aecheon Jung, Soyun Choi, Junhong Min et al.

Image segmentation is a vital task for providing human assistance and enhancing autonomy in our daily lives. In particular, RGB-D segmentation-leveraging both visual and depth cues-has attracted increasing attention as it promises richer scene understanding than RGB-only methods. However, most existing efforts have primarily focused on semantic segmentation and thus leave a critical gap. There is a relative scarcity of instance-level RGB-D segmentation datasets, which restricts current methods to broad category distinctions rather than fully capturing the fine-grained details required for recognizing individual objects. To bridge this gap, we introduce three RGB-D instance segmentation benchmarks, distinguished at the instance level. These datasets are versatile, supporting a wide range of applications from indoor navigation to robotic manipulation. In addition, we present an extensive evaluation of various baseline models on these benchmarks. This comprehensive analysis identifies both their strengths and shortcomings, guiding future work toward more robust, generalizable solutions. Finally, we propose a simple yet effective method for RGB-D data integration. Extensive evaluations affirm the effectiveness of our approach, offering a robust framework for advancing toward more nuanced scene understanding.

Weizheng Wang, Baijian Yang, Sungeun Hong et al.

Predicting crowd intentions and trajectories is critical for a range of real-world applications, involving social robotics and autonomous driving. Accurately modeling such behavior remains challenging due to the complexity of pairwise spatial-temporal interactions and the heterogeneous influence of groupwise dynamics. To address these challenges, we propose Hyper-STTN, a Hypergraph-based Spatial-Temporal Transformer Network for crowd trajectory prediction. Hyper-STTN constructs multiscale hypergraphs of varying group sizes to model groupwise correlations, captured through spectral hypergraph convolution based on random-walk probabilities. In parallel, a spatial-temporal transformer is employed to learn pedestrians' pairwise latent interactions across spatial and temporal dimensions. These heterogeneous groupwise and pairwise features are subsequently fused and aligned via a multimodal transformer. Extensive experiments on public pedestrian motion datasets demonstrate that Hyper-STTN consistently outperforms state-of-the-art baselines and ablation models.

Youngeun Kim, Sungeun Hong, Seunghan Yang et al.

Partial Adaptation (PDA) addresses a practical scenario in which the target domain contains only a subset of classes in the source domain. While PDA should take into account both class-level and sample-level to mitigate negative transfer, current approaches mostly rely on only one of them. In this paper, we propose a novel approach to fully exploit multi-level associations that can arise in PDA. Our Associative Partial Domain Adaptation (APDA) utilizes intra-domain association to actively select out non-trivial anomaly samples in each source-private class that sample-level weighting cannot handle. Additionally, our method considers inter-domain association to encourage positive transfer by mapping between nearby target samples and source samples with high label-commonness. For this, we exploit feature propagation in a proposed label space consisting of source ground-truth labels and target probabilistic labels. We further propose a geometric guidance loss based on the label commonness of each source class to encourage positive transfer. Our APDA consistently achieves state-of-the-art performance across public datasets.

Donghyeon Cho, Sungeun Hong, Sungil Kang et al.

This paper proposes key instance selection based on video saliency covering objectness and dynamics for unsupervised video object segmentation (UVOS). Our method takes frames sequentially and extracts object proposals with corresponding masks for each frame. We link objects according to their similarity until the M-th frame and then assign them unique IDs (i.e., instances). Similarity measure takes into account multiple properties such as ReID descriptor, expected trajectory, and semantic co-segmentation result. After M-th frame, we select K IDs based on video saliency and frequency of appearance; then only these key IDs are tracked through the remaining frames. Thanks to these technical contributions, our results are ranked third on the leaderboard of UVOS DAVIS challenge.

Sungeun Hong, Wonjin Jung, Ilsang Woo et al.

Over the past decade, there has been a growing interest in human pose estimation. Although much work has been done on 2D pose estimation, 3D pose estimation has still been relatively studied less. In this paper, we propose a top-bottom based two-stage 3D estimation framework. GloabalNet and RefineNet in our 2D pose estimation process enable us to find occluded or invisible 2D joints while 2D-to-3D pose estimator composed of residual blocks is used to lift 2D joints to 3D joints effectively. The proposed method achieves promising results with mean per joint position error at 42.39 on the validation dataset on `3D Human Pose Estimation within the ECCV 2018 PoseTrack Challenge.'

Jaehong Kim, Sungeun Hong, Yongseok Choi et al.

We propose doubly nested network(DNNet) where all neurons represent their own sub-models that solve the same task. Every sub-model is nested both layer-wise and channel-wise. While nesting sub-models layer-wise is straight-forward with deep-supervision as proposed in \cite{xie2015holistically}, channel-wise nesting has not been explored in the literature to our best knowledge. Channel-wise nesting is non-trivial as neurons between consecutive layers are all connected to each other. In this work, we introduce a technique to solve this problem by sorting channels topologically and connecting neurons accordingly. For the purpose, channel-causal convolutions are used. Slicing doubly nested network gives a working sub-network. The most notable application of our proposed network structure with slicing operation is resource-efficient inference. At test time, computing resources such as time and memory available for running the prediction algorithm can significantly vary across devices and applications. Given a budget constraint, we can slice the network accordingly and use a sub-model for inference within budget, requiring no additional computation such as training or fine-tuning after deployment. We demonstrate the effectiveness of our approach in several practical scenarios of utilizing available resource efficiently.

Sungeun Hong, Jongbin Ryu, Woobin Im et al.

Recognizing dynamic scenes is one of the fundamental problems in scene understanding, which categorizes moving scenes such as a forest fire, landslide, or avalanche. While existing methods focus on reliable capturing of static and dynamic information, few works have explored frame selection from a dynamic scene sequence. In this paper, we propose dynamic scene recognition using a deep dual descriptor based on `key frames' and `key segments.' Key frames that reflect the feature distribution of the sequence with a small number are used for capturing salient static appearances. Key segments, which are captured from the area around each key frame, provide an additional discriminative power by dynamic patterns within short time intervals. To this end, two types of transferred convolutional neural network features are used in our approach. A fully connected layer is used to select the key frames and key segments, while the convolutional layer is used to describe them. We conducted experiments using public datasets as well as a new dataset comprised of 23 dynamic scene classes with 10 videos per class. The evaluation results demonstrated the state-of-the-art performance of the proposed method.