Seunghyuk Cho

Seunghyuk Cho, Juyong Lee, Jaesik Park et al.

We present a rotated hyperbolic wrapped normal distribution (RoWN), a simple yet effective alteration of a hyperbolic wrapped normal distribution (HWN). The HWN expands the domain of probabilistic modeling from Euclidean to hyperbolic space, where a tree can be embedded with arbitrary low distortion in theory. In this work, we analyze the geometric properties of the diagonal HWN, a standard choice of distribution in probabilistic modeling. The analysis shows that the distribution is inappropriate to represent the data points at the same hierarchy level through their angular distance with the same norm in the Poincaré disk model. We then empirically verify the presence of limitations of HWN, and show how RoWN, the proposed distribution, can alleviate the limitations on various hierarchical datasets, including noisy synthetic binary tree, WordNet, and Atari 2600 Breakout. The code is available at https://github.com/ml-postech/RoWN.

Seunghyuk Cho, Sunghyun Choi, Jaeseung Heo et al.

Multi-agent systems (MAS) have substantially advanced autonomous software engineering (SWE), but their growing inference energy demands raise sustainability concerns. In this paper, we demonstrate that this cost is concentrated in an overlooked source: redundant output tokens generated across agents. Two empirical findings ground this claim. First, our per-token energy attribution for MAS reveals a sharp asymmetry: an output token consumes 30 to 1,000 times more energy than an input or cached token. Second, MAS inflate per-episode output because agents repeatedly re-explore overlapping repository regions. To address this inefficiency, we propose Librarian, a persistent search sub-agent that tracks repository-search history and suppresses redundant exploration actions across agents. By returning short references to file regions instead of full file excerpts, Librarian further reduces output-token volume. On SWE-Bench Verified, Librarian reduces per-episode GPU energy consumption of existing multi-agent SWE systems by up to 25% while preserving task performance.

Seunghyuk Cho, Juyong Lee, Dongwoo Kim

We propose a Gaussian manifold variational auto-encoder (GM-VAE) whose latent space consists of a set of Gaussian distributions. It is known that the set of the univariate Gaussian distributions with the Fisher information metric form a hyperbolic space, which we call a Gaussian manifold. To learn the VAE endowed with the Gaussian manifolds, we propose a pseudo-Gaussian manifold normal distribution based on the Kullback-Leibler divergence, a local approximation of the squared Fisher-Rao distance, to define a density over the latent space. In experiments, we demonstrate the efficacy of GM-VAE on two different tasks: density estimation of image datasets and environment modeling in model-based reinforcement learning. GM-VAE outperforms the other variants of hyperbolic- and Euclidean-VAEs on density estimation tasks and shows competitive performance in model-based reinforcement learning. We observe that our model provides strong numerical stability, addressing a common limitation reported in previous hyperbolic-VAEs.

Sungjun Cho, Seunghyuk Cho, Sungwoo Park et al.

Real-world graphs naturally exhibit hierarchical or cyclical structures that are unfit for the typical Euclidean space. While there exist graph neural networks that leverage hyperbolic or spherical spaces to learn representations that embed such structures more accurately, these methods are confined under the message-passing paradigm, making the models vulnerable against side-effects such as oversmoothing and oversquashing. More recent work have proposed global attention-based graph Transformers that can easily model long-range interactions, but their extensions towards non-Euclidean geometry are yet unexplored. To bridge this gap, we propose Fully Product-Stereographic Transformer, a generalization of Transformers towards operating entirely on the product of constant curvature spaces. When combined with tokenized graph Transformers, our model can learn the curvature appropriate for the input graph in an end-to-end fashion, without the need of additional tuning on different curvature initializations. We also provide a kernelized approach to non-Euclidean attention, which enables our model to run in time and memory cost linear to the number of nodes and edges while respecting the underlying geometry. Experiments on graph reconstruction and node classification demonstrate the benefits of generalizing Transformers to the non-Euclidean domain.

MoonJeong Park, Seungbeom Lee, Kyungmin Kim et al.

Many existing transductive bounds rely on classical complexity measures that are computationally intractable and often misaligned with empirical behavior. In this work, we establish new representation-based generalization bounds in a distribution-free transductive setting, where learned representations are dependent, and test features are accessible during training. We derive global and class-wise bounds via optimal transport, expressed in terms of Wasserstein distances between encoded feature distributions. We demonstrate that our bounds are efficiently computable and strongly correlate with empirical generalization in graph node classification, improving upon classical complexity measures. Additionally, our analysis reveals how the GNN aggregation process transforms the representation distributions, inducing a trade-off between intra-class concentration and inter-class separation. This yields depth-dependent characterizations that capture the non-monotonic relationship between depth and generalization error observed in practice. The code is available at https://github.com/ml-postech/Transductive-OT-Gen-Bound.

Saemi Moon, Seunghyuk Cho, Dongwoo Kim

We tackle the problem of feature unlearning from a pre-trained image generative model: GANs and VAEs. Unlike a common unlearning task where an unlearning target is a subset of the training set, we aim to unlearn a specific feature, such as hairstyle from facial images, from the pre-trained generative models. As the target feature is only presented in a local region of an image, unlearning the entire image from the pre-trained model may result in losing other details in the remaining region of the image. To specify which features to unlearn, we collect randomly generated images that contain the target features. We then identify a latent representation corresponding to the target feature and then use the representation to fine-tune the pre-trained model. Through experiments on MNIST, CelebA, and FFHQ datasets, we show that target features are successfully removed while keeping the fidelity of the original models. Further experiments with an adversarial attack show that the unlearned model is more robust under the presence of malicious parties.

Seunghyuk Cho, Zhenyue Qin, Yang Liu et al.

We introduce GeoDANO, a geometric vision-language model (VLM) with a domain-agnostic vision encoder, for solving plane geometry problems. Although VLMs have been employed for solving geometry problems, their ability to recognize geometric features remains insufficiently analyzed. To address this gap, we propose a benchmark that evaluates the recognition of visual geometric features, including primitives such as dots and lines, and relations such as orthogonality. Our preliminary study shows that vision encoders often used in general-purpose VLMs, e.g., OpenCLIP, fail to detect these features and struggle to generalize across domains. To overcome the limitation, we develop GeoCLIP, a CLIP-based model trained on synthetic geometric diagram--caption pairs. Benchmark results show that GeoCLIP outperforms existing vision encoders in recognizing geometric features. We then propose our VLM, GeoDANO, which augments GeoCLIP with a domain adaptation strategy for unseen diagram styles. GeoDANO outperforms specialized methods for plane geometry problems and GPT-4o on MathVerse. The implementation is available at https://github.com/ml-postech/GeoDANO.

Youngbin Choi, Seunghyuk Cho, Minjong Lee et al.

Personalizing large language models (LLMs) is important for aligning outputs with diverse user preferences, yet existing methods struggle with flexibility and generalization. We propose CoPL (Collaborative Preference Learning), a graph-based collaborative filtering framework that models user-response relationships to enhance preference estimation, particularly in sparse annotation settings. By integrating a mixture of LoRA experts, CoPL efficiently fine-tunes LLMs while dynamically balancing shared and user-specific preferences. Additionally, an optimization-free adaptation strategy enables generalization to unseen users without fine-tuning. Experiments on UltraFeedback-P demonstrate that CoPL outperforms existing personalized reward models, effectively capturing both common and controversial preferences, making it a scalable solution for personalized LLM alignment. The code is available at https://github.com/ml-postech/CoPL.

Juyong Lee, Seunghyuk Cho

Consistency training, which exploits both supervised and unsupervised learning with different augmentations on image, is an effective method of utilizing unlabeled data in semi-supervised learning (SSL) manner. Here, we present another version of the method with Grad-CAM consistency loss, so it can be utilized in training model with better generalization and adjustability. We show that our method improved the baseline ResNet model with at most 1.44 % and 0.31 $\pm$ 0.59 %p accuracy improvement on average with CIFAR-10 dataset. We conducted ablation study comparing to using only psuedo-label for consistency training. Also, we argue that our method can adjust in different environments when targeted to different units in the model. The code is available: https://github.com/gimme1dollar/gradcam-consistency-semi-sup.

Seokwon Yoon, Youngbin Choi, Seunghyuk Cho et al.

Generative Flow Networks (GFlowNets) learn to sample diverse candidates in proportion to a reward function, making them well-suited for scientific discovery, where exploring multiple promising solutions is crucial. Further extending GFlowNets to multi-objective settings has attracted growing interest since real-world applications often involve multiple, conflicting objectives. However, existing approaches require additional training for each set of objectives, limiting their applicability and incurring substantial computational overhead. We propose a training-free mixing policy that composes pre-trained GFlowNets at inference time, enabling rapid adaptation without finetuning or retraining. Importantly, our framework is flexible, capable of handling diverse reward combinations ranging from linear scalarization to complex non-linear logical operators, which are often handled separately in previous literature. We prove that our method exactly recovers the target distribution for linear scalarization and quantify the approximation quality for nonlinear operators through a distortion factor. Experiments on a synthetic 2D grid and real-world molecule-generation tasks demonstrate that our approach achieves performance comparable to baselines that require additional training.

Seunghyuk Cho, Zhenyue Qin, Yang Liu et al.

Plane geometry problem solving (PGPS) has recently gained significant attention as a benchmark to assess the multi-modal reasoning capabilities of large vision-language models. Despite the growing interest in PGPS, the research community still lacks a comprehensive overview that systematically synthesizes recent work in PGPS. To fill this gap, we present a survey of existing PGPS studies. We first categorize PGPS methods into an encoder-decoder framework and summarize the corresponding output formats used by their encoders and decoders. Subsequently, we classify and analyze these encoders and decoders according to their architectural designs. Finally, we outline major challenges and promising directions for future research. In particular, we discuss the hallucination issues arising during the encoding phase within encoder-decoder architectures, as well as the problem of data leakage in current PGPS benchmarks.

Youngbin Choi, Minjong Lee, Saemi Moon et al.

Large language models (LLMs) are increasingly studied in the context of multi-turn reasoning, where models iteratively refine their outputs based on user-provided feedback. Such settings are crucial for tasks that require complex reasoning, yet existing feedback paradigms often rely on issuing new messages. LLMs struggle to integrate these reliably, leading to inconsistent improvements. In this work, we introduce in-place feedback, a novel interaction paradigm in which users directly edit an LLM's previous response, and the model conditions on this modified response to generate its revision. Empirical evaluations on diverse reasoning-intensive benchmarks reveal that in-place feedback achieves better performance than conventional multi-turn feedback while using $79.1\%$ fewer tokens. Complementary analyses on controlled environments further demonstrate that in-place feedback resolves a core limitation of multi-turn feedback: models often fail to apply feedback precisely to erroneous parts of the response, leaving errors uncorrected and sometimes introducing new mistakes into previously correct content. These findings suggest that in-place feedback offers a more natural and effective mechanism for guiding LLMs in reasoning-intensive tasks.

Hoyoung Kim, Seunghyuk Cho, Dongwoo Kim et al.

Crowdsourcing systems enable us to collect large-scale dataset, but inherently suffer from noisy labels of low-paid workers. We address the inference and learning problems using such a crowdsourced dataset with noise. Due to the nature of sparsity in crowdsourcing, it is critical to exploit both probabilistic model to capture worker prior and neural network to extract task feature despite risks from wrong prior and overfitted feature in practice. We hence establish a neural-powered Bayesian framework, from which we devise deepMF and deepBP with different choice of variational approximation methods, mean field (MF) and belief propagation (BP), respectively. This provides a unified view of existing methods, which are special cases of deepMF with different priors. In addition, our empirical study suggests that deepBP is a new approach, which is more robust against wrong prior, feature overfitting and extreme workers thanks to the more sophisticated BP than MF.

Seunghyuk Cho, Minsoo Koo, Dongwoo Kim et al.

With rise of interventional cardiology, Catheter Ablation Therapy (CAT) has established itself as a first-line solution to treat cardiac arrhythmia. Although CAT is a promising technique, cardiologist lacks vision inside the body during the procedure, which may cause serious clinical syndromes. To support accurate clinical procedure, Contact Force Sensing (CFS) system is developed to find a position of the catheter tip through the measure of contact force between catheter and heart tissue. However, the practical usability of commercialized CFS systems is not fully understood due to inaccuracy in the measurement. To support the development of more accurate system, we develop a full pipeline of CFS system with newly collected benchmark dataset through a contact force sensing catheter in simplest hardware form. Our dataset was roughly collected with human noise to increase data diversity. Through the analysis of the dataset, we identify a problem defined as Shift of Reference (SoR), which prevents accurate measurement of contact force. To overcome the problem, we conduct the contact force estimation via standard deep neural networks including for Recurrent Neural Network (RNN), Fully Convolutional Network (FCN) and Transformer. An average error in measurement for RNN, FCN and Transformer are, respectively, 2.46g, 3.03g and 3.01g. Through these studies, we try to lay a groundwork, serve a performance criteria for future CFS system research and open a publicly available dataset to public.