Sangwoo Seo

Publications

* indicates equal contribution.

1. ELASTIC: Trajectory-Based Synchronization of Event and Tracking Data in Soccer

   Hyunsung Kim, Hoyoung Choi, Kunhee Lee,  Sangwoo Seo, Tom Boomstra, Jinsung Yoon,  and Chanyoung Park

   ACM International Conference on Information and Knowledge Management (CIKM), 2026

   Abstract Paper Code

   Combining event and tracking data is fundamental to modern soccer analytics, yet the two sources are rarely well aligned: event timestamps recorded by human annotators often miss the true moment of the action, distorting the spatiotemporal context that downstream models rely on. Existing synchronization methods depend on noisy human-annotated event locations and fail to detect ball receptions, obscuring when each player gains ball possession. To address these limitations, we propose ELASTIC (Event-Location-AgnoSTIC synchronizer), a framework that infers the start and end timestamps of events solely from player and ball trajectories, without relying on annotated event locations. To recover ball receptions, ELASTIC enriches the event sequence by inserting virtual termination events between consecutive events, so that the end of each event is detected jointly with its start. It then extracts a sparse set of candidate frames where ball touches are physically plausible, and aligns the termination-inserted event sequence with the candidate-frame sequence using an extended Needleman-Wunsch algorithm. For reproducible evaluation, we construct a publicly available benchmark by annotating ground-truth timestamps on the Sportec Open DFL Dataset, on which ELASTIC substantially outperforms existing methods. Through downstream task evaluation, we further show that improved synchronization translates into measurable gains in soccer analytics.

2. Bridging Circuit Graphs and Language: A Cross-Modal Transformer for Netlist-to-Text Alignment and Generation

   Sangwoo Seo, Hyunsung Kim, Hyunchul Kim, Byeonggon Kang, Rex Ying,  and Chanyoung Park

   ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) Workshop on Graph Machine Learning for the Large Model Era, 2026

   Abstract

   As integrated circuit (IC) design grows in complexity, foundation models for circuits have emerged in two forms: encoders for predictive tasks and decoders for generative tasks. While LLM-based circuit decoders excel at producing circuit-related text, their reliance on a text-based latent space limits their ability to capture structural information from the circuit-graph latent space. Recent work bridges this gap using an MLP-based connector trained with autoregressive loss, but struggles to align complex netlists with linguistically contextualized text and tends to prioritize next-token prediction over summarizing critical netlist information. We propose NetlistFormer, a cross-modal circuit transformer for netlist-to-text alignment and generation, which uses learnable query vectors over the netlist graph in a two-stage framework. In the first stage, the queries act as functional sub-structure selectors that extract and compress key structural information from the netlist graph under textual guidance, via a multi-objective cross-modal alignment beyond autoregressive loss. In the second stage, the resulting queries are passed to an LLM for netlist-language generation. Experiments show that NetlistFormer consistently outperforms general-purpose LLMs and MLP-based connector models across standard similarity metrics and domain-aware keyword-level metrics. To our knowledge, NetlistFormer is the first circuit generative transformer trained with multi-objective cross-modal alignment.

3. Adaptive Graph Rewiring to Mitigate Over-Squashing in Mesh-Based GNNs for Fluid Dynamics Simulations

   Sangwoo Seo, Hyunsung Kim, Jiwan Kim,  and Chanyoung Park

   ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) Workshop on Reliable Scientific Foundation Models, 2026

   Abstract Paper

   Mesh-based simulation using Graph Neural Networks (GNNs) has been recognized as a promising approach for modeling fluid dynamics. However, the mesh refinement techniques which allocate finer resolution to regions with steep gradients can induce the over-squashing problem in mesh-based GNNs, which prevents the capture of long-range physical interactions. Conventional graph rewiring methods attempt to alleviate this issue by adding new edges, but they typically complete all rewiring operations before applying them to the GNN. These approaches are physically unrealistic, as they assume instantaneous interactions between distant nodes and disregard the distance information between particles. To address these limitations, we propose a novel framework, called Adaptive Graph Rewiring in Mesh-Based Graph Neural Networks (AdaMeshNet), that introduces an adaptive rewiring process into the message-passing procedure to model the gradual propagation of physical interactions. Our method computes a rewiring delay score for bottleneck nodes in the mesh graph, based on the shortest-path distance and the velocity difference. Using this score, it dynamically selects the message-passing layer at which new edges are rewired, which can lead to adaptive rewiring in a mesh graph. Extensive experiments on mesh-based fluid simulations demonstrate that AdaMeshNet outperforms conventional rewiring methods, effectively modeling the sequential nature of physical interactions and enabling more accurate predictions.

4. PathCRF: Ball-Free Soccer Event Detection via Possession Path Inference from Player Trajectories

   Hyunsung Kim, Kunhee Lee,  Sangwoo Seo, Sangki Ko, Jinsung Yoon,  and Chanyoung Park

   ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), 2026

   Abstract Paper Code

   Despite recent advances in AI, event data collection in soccer still relies heavily on labor-intensive manual annotation. Although prior work has explored automatic event detection using player and ball trajectories, ball tracking also remains difficult to scale due to high infrastructural and operational costs. As a result, comprehensive data collection in soccer is largely confined to top-tier competitions, limiting the broader adoption of data-driven analysis in this domain. To address this challenge, this paper proposes PathCRF, a framework for detecting on-ball soccer events using only player tracking data. We model player trajectories as a fully connected dynamic graph and formulate event detection as the problem of selecting exactly one edge corresponding to the current possession state at each time step. To ensure logical consistency of the resulting edge sequence, we employ a Conditional Random Field (CRF) that forbids impossible transitions between consecutive edges, where emission and transition scores are dynamically computed from edge embeddings produced by a socio-temporal backbone architecture. During inference, the most probable edge sequence is obtained via Viterbi decoding, and events such as ball controls or passes are detected whenever the selected edge changes between adjacent time steps. Experiments show that PathCRF produces accurate, logically consistent possession paths, enabling reliable downstream analyses while substantially reducing the need for manual event annotation.

5. CompoDistill: Attention Distillation for Compositional Reasoning in Multimodal LLMs

   Jiwan Kim, Kibum Kim,  Sangwoo Seo,  and Chanyoung Park

   International Conference on Learning Representations (ICLR), 2026

   Abstract Paper Code

   Recently, efficient Multimodal Large Language Models (MLLMs) have gained significant attention as a solution to their high computational complexity, making them more practical for real-world applications. In this regard, the knowledge distillation (KD) approach has emerged as a promising alternative, which transfers the rich visual and linguistic knowledge from a larger model (teacher) to a smaller model (student). However, we observe that existing KD methods struggle to effectively distill the teacher MLLM’s rich visual perception abilities to the student, a challenge that has been largely overlooked in previous studies. Through a systematic analysis, we identify visual attention misalignment between student and teacher as the main cause of this issue. Based on this insight, we propose CompoDistill, a novel KD framework that explicitly aligns the student’s visual attention with that of the teacher to enhance the student’s visual perception abilities. Our extensive experiments show that CompoDistill significantly improves performance on compositional reasoning tasks that require visual perception abilities while maintaining strong performance on visual question answering tasks, as done in existing studies. Furthermore, CompoDistill demonstrates effectiveness with a more advanced backbone, highlighting its generalizability.

6. Better Prevent than Tackle: Valuing Defense in Soccer Based on Graph Neural Networks

   Hyunsung Kim,  Sangwoo Seo, Hoyoung Choi, Tom Boomstra, Jinsung Yoon,  and Chanyoung Park

   MIT Sloan Sports Analytics Conference (SSAC), 2026

   Abstract Paper Code

   Evaluating defensive performance in soccer remains challenging, as effective defending is often expressed not through visible on-ball actions such as interceptions and tackles, but through preventing dangerous opportunities before they arise. Existing approaches have largely focused on valuing on-ball actions, leaving much of defenders’ true impact unmeasured. To address this gap, we propose DEFCON (DEFensive CONtribution evaluator), a comprehensive framework that quantifies player-level defensive contributions for every attacking situation in soccer. Leveraging Graph Attention Networks, DEFCON estimates the success probability and expected value of each attacking option, along with each defender’s responsibility for stopping it. These components yield an Expected Possession Value (EPV) for the attacking team before and after each action, and DEFCON assigns positive or negative credits to defenders according to whether they reduced or increased the opponent’s EPV. Trained on 2023-24 and evaluated on 2024-25 Eredivisie event and tracking data, DEFCON’s aggregated player credits exhibit strong positive correlations with market valuations. Finally, we showcase several practical applications, including in-game timelines of defensive contributions, spatial analyses across pitch zones, and pairwise summaries of attacker-defender interactions.

7. Target Circuit Matching in Large-Scale Netlists using GNN-Based Region Prediction

   Sangwoo Seo, Jimin Seo, Yoonho Lee, Donghyeon Kim, Hyejin Shin, Banghyun Sung,  and Chanyoung Park

   International Conference on Computer-Aided Design (ICCAD), 2025

   Abstract Paper Code

   Subgraph matching plays an important role in electronic design automation (EDA) and circuit verification. Traditional rule-based methods have limitations in generalizing to arbitrary target circuits. Furthermore, node-to-node matching approaches tend to be computationally inefficient, particularly for large-scale circuits. Deep learning methods have emerged as a potential solution to address these challenges, but existing models fail to efficiently capture global subgraph embeddings or rely on inefficient matching matrices, which limits their effectiveness for large circuits. In this paper, we propose an efficient graph matching approach that utilizes Graph Neural Networks (GNNs) to predict regions of high probability for containing the target circuit. Specifically, we construct various negative samples to enable GNNs to accurately learn the presence of target circuits and develop an approach to directly extracting subgraph embeddings from the entire circuit, which captures global subgraph information and addresses the inefficiency of applying GNNs to all candidate subgraphs. Extensive experiments demonstrate that our approach significantly outperforms existing methods in terms of time efficiency and target region prediction, offering a scalable and effective solution for subgraph matching in large-scale circuits.

8. Disentangling Hyperedges through the Lens of Category Theory

   Yoonho Lee, Junseok Lee,  Sangwoo Seo, Sungwon Kim, Yeongmin Kim,  and Chanyoung Park

   Conference on Neural Information Processing Systems (NeurIPS), 2025

   Abstract Paper Code

   Despite the promising results of disentangled representation learning in discovering latent patterns in graph-structured data, few studies have explored disentanglement for hypergraph-structured data. Integrating hyperedge disentanglement into hypergraph neural networks enables models to leverage hidden hyperedge semantics, such as unannotated relations between nodes, that are associated with labels. This paper presents an analysis of hyperedge disentanglement from a category-theoretical perspective and proposes a novel criterion for disentanglement derived from the naturality condition. Our proof-of-concept model experimentally showed the potential of the proposed criterion by successfully capturing functional relations of genes (nodes) in genetic pathways (hyperedges).

9. ELASTIC: Event-Tracking Data Synchronization in Soccer Without Annotated Event Locations

   Hyunsung Kim, Hoyoung Choi,  Sangwoo Seo, Tom Boomstra, Jinsung Yoon,  and Chanyoung Park

   European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD) Workshop on MLSA, 2025

   Abstract Paper Code

   The integration of event and tracking data has become essential for advanced analysis in soccer. However, synchronizing these two modalities remains a significant challenge due to temporal and spatial inaccuracies in manually recorded event timestamps. Existing synchronizers typically rely on annotated event locations, which themselves are prone to spatial errors and thus can distort synchronization results. To address this issue, we propose ELASTIC (Event-Location-AgnoSTIC synchronizer), a synchronization framework that only uses features derived from tracking data. ELASTIC also explicitly detects the end times of pass-like events and separates the detection of major and minor events, which improves the completeness of the synchronized output and reduces error cascade across events. We annotated the ground truth timestamps of 2,134 events from three Eredivisie matches to measure the synchronization accuracy, and the experimental results demonstrate that ELASTIC outperforms existing synchronizers by a large margin.

10. Global Context-aware Representation Learning for Spatially Resolved Transcriptomics

    Yunhak Oh*, Junseok Lee*, Yeongmin Kim,  Sangwoo Seo, Namkyeong Lee,  and Chanyoung Park

    International Conference on Machine Learning (ICML), 2025

    Abstract Paper Code

    Spatially Resolved Transcriptomics (SRT) is a cutting-edge technique that captures the spatial context of cells within tissues, enabling the study of complex biological networks. Recent graph-based methods leverage both gene expression and spatial information to identify relevant spatial domains. However, these approaches fall short in obtaining meaningful spot representations, especially for spots near spatial domain boundaries, as they heavily emphasize adjacent spots that have minimal feature differences from an anchor node. To address this, we propose Spotscape, a novel framework that introduces the Similarity Telescope module to capture global relationships between multiple spots. Additionally, we propose a similarity scaling strategy to regulate the distances between intra- and inter-slice spots, facilitating effective multi-slice integration. Extensive experiments demonstrate the superiority of Spotscape in various downstream tasks, including single-slice and multi-slice scenarios.

11. Capturing Functional Context of Genetic Pathways through Hyperedge Disentanglement

    Yoonho Lee, Junseok Lee,  Sangwoo Seo, Sungwon Kim, Yeongmin Kim,  and Chanyoung Park

    International Conference on Learning Representations (ICLR) Workshop on Machine Learning for Genomics Explorations, 2025

    Abstract Code

    The hypergraph data structure has been used to represent the multiway interactions of a set of genes of a genetic pathway. Since genes within each genetic pathway collaboratively perform a biological function, the functional context of a pathway (i.e., the interaction context of a hyperedge), which is often unannotated, needs to be captured. However, most existing hypergraph neural networks fail to reflect the interaction context of each hyperedge due to their limited ability to capture important or relevant factors. In this paper, we propose a simple yet effective hyperedge disentangling method, Natural-HNN, which captures the interaction context of a hyperedge. We introduce a novel guidance mechanism for hyperedge disentanglement based on the naturality condition in category theory. In our experiments, we applied our model to hypergraphs of genetic pathways for the cancer subtype classification task and demonstrated that our model outperforms baseline approaches by capturing the functional semantic similarity of genetic pathways.

12. Self-Explainable Temporal Graph Networks based on Graph Information Bottleneck

    Sangwoo Seo, Sungwon Kim, Jihyeong Jung, Yoonho Lee,  and Chanyoung Park

    ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), 2024

    Abstract Paper Code

    Temporal Graph Neural Networks (TGNN) have the ability to capture both the graph topology and dynamic dependencies of interactions within a graph over time. There has been a growing need to explain the predictions of TGNN models due to the difficulty in identifying how past events influence their predictions. Since the explanation model for a static graph cannot be readily applied to temporal graphs due to its inability to capture temporal dependencies, recent studies proposed explanation models for temporal graphs. However, existing explanation models for temporal graphs rely on post-hoc explanations, requiring separate models for prediction and explanation, which is limited in two aspects: efficiency and accuracy of explanation. In this work, we propose a novel built-in explanation framework for temporal graphs, called Self-Explainable Temporal Graph Networks based on Graph Information Bottleneck (TGIB). TGIB provides explanations for event occurrences by introducing stochasticity in each temporal event based on the Information Bottleneck theory. Experimental results demonstrate the superiority of TGIB in terms of both the link prediction performance and explainability compared to state-of-the-art methods. This is the first work that simultaneously performs prediction and explanation for temporal graphs in an end-to-end manner.

13. Interpretable Graph Model with Prototype-Based Graph Information Bottleneck

    Sangwoo Seo, Sungwon Kim,  and Chanyoung Park

    ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) Workshop on Human-Interpretable AI, 2024

    Best Paper Award

    Abstract Paper

    The success of Graph Neural Networks (GNNs) has led to a need for understanding their decision-making process and providing explanations for their predictions, which has given rise to explainable AI (XAI) that offers transparent explanations for black-box models. Recently, the use of prototypes has successfully improved the explainability of models by learning prototypes to imply training graphs that affect the prediction. However, these approaches tend to provide prototypes with excessive information from the entire graph, leading to the exclusion of key substructures or the inclusion of irrelevant substructures, which can limit both the interpretability and the performance of the model in downstream tasks. In this work, we propose a novel framework of explainable GNNs, called interpretable Prototype-based Graph Information Bottleneck (PGIB) that incorporates prototype learning within the information bottleneck framework to provide prototypes with the key subgraph from the input graph that is important for the model prediction. This is the first work that incorporates prototype learning into the process of identifying the key subgraphs that have a critical impact on the prediction performance. Extensive experiments, including qualitative analysis, demonstrate that PGIB outperforms state-of-the-art methods in terms of both prediction performance and explainability.

14. Unsupervised Episode Generation for Graph Meta-learning

    Jihyeong Jung,  Sangwoo Seo, Sungwon Kim,  and Chanyoung Park

    International Conference on Machine Learning (ICML), 2024

    Abstract Paper Code

    We propose Unsupervised Episode Generation method called Neighbors as Queries (NaQ) to solve the Few-Shot Node-Classification (FSNC) task by unsupervised Graph Meta-learning. Doing so enables full utilization of the information of all nodes in a graph, which is not possible in current supervised meta-learning methods for FSNC due to the label-scarcity problem. In addition, unlike unsupervised Graph Contrastive Learning (GCL) methods that overlook the downstream task to be solved at the training phase resulting in vulnerability to class imbalance of a graph, we adopt the episodic learning framework that allows the model to be aware of the downstream task format, i.e., FSNC. The proposed NaQ is a simple but effective unsupervised episode generation method that randomly samples nodes from a graph to make a support set, followed by similarity-based sampling of nodes to make the corresponding query set. Since NaQ is model-agnostic, any existing supervised graph meta-learning methods can be trained in an unsupervised manner, while not sacrificing much of their performance or sometimes even improving them. Extensive experimental results demonstrate the effectiveness of our proposed unsupervised episode generation method for graph meta-learning towards the FSNC task.

15. Interpretable Prototype-based Graph Information Bottleneck

    Sangwoo Seo, Sungwon Kim,  and Chanyoung Park

    Conference on Neural Information Processing Systems (NeurIPS), 2023

    Gold Prize at the 30th Samsung HumanTech Paper Award

    Abstract Paper Code

    The success of Graph Neural Networks (GNNs) has led to a need for understanding their decision-making process and providing explanations for their predictions, which has given rise to explainable AI (XAI) that offers transparent explanations for black-box models. Recently, the use of prototypes has successfully improved the explainability of models by learning prototypes to imply training graphs that affect the prediction. However, these approaches tend to provide prototypes with excessive information from the entire graph, leading to the exclusion of key substructures or the inclusion of irrelevant substructures, which can limit both the interpretability and the performance of the model in downstream tasks. In this work, we propose a novel framework of explainable GNNs, called interpretable Prototype-based Graph Information Bottleneck (PGIB) that incorporates prototype learning within the information bottleneck framework to provide prototypes with the key subgraph from the input graph that is important for the model prediction. This is the first work that incorporates prototype learning into the process of identifying the key subgraphs that have a critical impact on the prediction performance. Extensive experiments, including qualitative analysis, demonstrate that PGIB outperforms state-of-the-art methods in terms of both prediction performance and explainability.

16. Explanation-based Graph Neural Networks for Graph Classification

    Sangwoo Seo, Seungjun Jung,  and Changick Kim

    International Conference on Pattern Recognition (ICPR), 2022

    Abstract Paper

    Graph Neural Network models can be used to quickly analyze interactions between multiple data expressed in a graph structure, with high accuracy. Previous studies accurately extract subgraphs which have a significant influence on the whole graph, providing accurate explanations for predictions of GNN. We noted that explanation components could help improve classification performance as unique representations of each class. Therefore, we suggest the GNN performance can be further improved by using explanation components. In this paper, we propose an Explanation-Based Graph Neural Networks (EBGNN) that utilizes contrastive learning at the instance level, by applying explanation components. In EBGNN, the explanation components ensure similarity for instances within the same class, and promote separability for instances in different classes. Finally, we conducted an evaluation on five benchmark datasets (MUTAG, IMDB-BINARY, PROTEINS, NCI1, and DD). Our experiment showed a significant increase in graph classification performance compared to state-of-the-art methods.