Taeyoung Kim

Taeyoung Kim, Joon-Hyuk Ko · eth-zurich

We propose an architecture that augments the Flux Neural Operator (Flux NO), which combines the classical finite volume method (FVM) with neural operators, with ViT-based context injection. Our model is formulated as a hypernetwork: it extracts solution dynamics over a finite temporal window, encodes them with a recurrent Vision Transformer, and generates the parameters of a context-conditioned neural operator. This enables the model to infer and solve conservation laws without explicit access to the governing equation or PDE coefficients. Experimentally, we show that the proposed method preserves the robustness, generalization ability, and long-time prediction advantages of Flux NO over standard neural operators, while delivering reliable numerical solutions across a broad range of conservative systems, including previously unseen fluxes. Our code is available at https://github.com/xx257xx/CONTEXT_FLUX_NO.

Injoon Cho, Praveen Kumar Rajendran, Taeyoung Kim et al.

As the demand for autonomous driving increases, it is paramount to ensure safety. Early accident prediction using deep learning methods for driving safety has recently gained much attention. In this task, early accident prediction and a point prediction of where the drivers should look are determined, with the dashcam video as input. We propose to exploit the double actors and regularized critics (DARC) method, for the first time, on this accident forecasting platform. We derive inspiration from DARC since it is currently a state-of-the-art reinforcement learning (RL) model on continuous action space suitable for accident anticipation. Results show that by utilizing DARC, we can make predictions 5\% earlier on average while improving in multiple metrics of precision compared to existing methods. The results imply that using our RL-based problem formulation could significantly increase the safety of autonomous driving.

Taeyoung Kim, Myungjoo Kang

A Fourier neural operator (FNO) is one of the physics-inspired machine learning methods. In particular, it is a neural operator. In recent times, several types of neural operators have been developed, e.g., deep operator networks, Graph neural operator (GNO), and Multiwavelet-based operator (MWTO). Compared with other models, the FNO is computationally efficient and can learn nonlinear operators between function spaces independent of a certain finite basis. In this study, we investigated the bounding of the Rademacher complexity of the FNO based on specific group norms. Using capacity based on these norms, we bound the generalization error of the model. In addition, we investigated the correlation between the empirical generalization error and the proposed capacity of FNO. From the perspective of our result, we inferred that the type of group norms determines the information about the weights and architecture of the FNO model stored in the capacity. And then, we confirmed these inferences through experiments. Based on this fact, we gained insight into the impact of the number of modes used in the FNO model on the generalization error. And we got experimental results that followed our insights.

Woohyeon Moon, Bumgeun Park, Sarvar Hussain Nengroo et al.

Recently, as the demand for cleaning robots has steadily increased, therefore household electricity consumption is also increasing. To solve this electricity consumption issue, the problem of efficient path planning for cleaning robot has become important and many studies have been conducted. However, most of them are about moving along a simple path segment, not about the whole path to clean all places. As the emerging deep learning technique, reinforcement learning (RL) has been adopted for cleaning robot. However, the models for RL operate only in a specific cleaning environment, not the various cleaning environment. The problem is that the models have to retrain whenever the cleaning environment changes. To solve this problem, the proximal policy optimization (PPO) algorithm is combined with an efficient path planning that operates in various cleaning environments, using transfer learning (TL), detection nearest cleaned tile, reward shaping, and making elite set methods. The proposed method is validated with an ablation study and comparison with conventional methods such as random and zigzag. The experimental results demonstrate that the proposed method achieves improved training performance and increased convergence speed over the original PPO. And it also demonstrates that this proposed method is better performance than conventional methods (random, zigzag).

Sumit Mishra, Medhavi Mishra, Taeyoung Kim et al.

Road infrastructure can affect the occurrence of road accidents. Therefore, identifying roadway features with high accident probability is crucial. Here, we introduce image inpainting that can assist authorities in achieving safe roadway design with minimal intervention in the current roadway structure. Image inpainting is based on inpainting safe roadway elements in a roadway image, replacing accident-prone (AP) features by using a diffusion model. After object-level segmentation, the AP features identified by the properties of accident hotspots are masked by a human operator and safe roadway elements are inpainted. With only an average time of 2 min for image inpainting, the likelihood of an image being classified as an accident hotspot drops by an average of 11.85%. In addition, safe urban spaces can be designed considering human factors of commuters such as gaze saliency. Considering this, we introduce saliency enhancement that suggests chrominance alteration for a safe road view.

Bumgeun Park, Taeyoung Kim, Woohyeon Moon et al.

Training agents via off-policy deep reinforcement learning (RL) requires a large memory, named replay memory, that stores past experiences used for learning. These experiences are sampled, uniformly or non-uniformly, to create the batches used for training. When calculating the loss function, off-policy algorithms assume that all samples are of the same importance. In this paper, we hypothesize that training can be enhanced by assigning different importance for each experience based on their temporal-difference (TD) error directly in the training objective. We propose a novel method that introduces a weighting factor for each experience when calculating the loss function at the learning stage. In addition to improving convergence speed when used with uniform sampling, the method can be combined with prioritization methods for non-uniform sampling. Combining the proposed method with prioritization methods improves sampling efficiency while increasing the performance of TD-based off-policy RL algorithms. The effectiveness of the proposed method is demonstrated by experiments in six environments of the OpenAI Gym suite. The experimental results demonstrate that the proposed method achieves a 33%~76% reduction of convergence speed in three environments and an 11% increase in returns and a 3%~10% increase in success rate for other three environments.

Dongyoung Kim, Sumin Park, Woomin Song et al.

Improving embodied reasoning in multimodal-large-language models (MLLMs) is essential for building vision-language-action models (VLAs) on top of them to readily translate multimodal understanding into low-level actions. Accordingly, recent work has explored enhancing embodied reasoning in MLLMs through supervision of vision-question-answering type. However, these approaches have been reported to result in unstable VLA performance, often yielding only marginal or even negative gains. In this paper, we propose a more systematic MLLM training framework RoboAlign that reliably improves VLA performance. Our key idea is to sample action tokens via zero-shot natural language reasoning and refines this reasoning using reinforcement learning (RL) to improve action accuracy. As a result, RoboAlign bridges the modality gap between language and low-level actions in MLLMs, and facilitate knowledge transfer from MLLM to VLA. To validate the effectiveness of RoboAlign, we train VLAs by adding a diffusion-based action head on top of an MLLM backbone and evaluate them on major robotics benchmarks. Remarkably, by performing RL-based alignment after SFT using less than 1\% of the data, RoboAlign achieves performance improvements of 17.5\%, 18.9\%, and 106.6\% over SFT baselines on LIBERO, CALVIN, and real-world environments, respectively.

Bumgeun Park, Taeyoung Kim, Quoc-Vinh Lai-Dang et al.

Efficient exploration for an agent is challenging in reinforcement learning (RL). In this paper, a novel actor-critic framework namely virtual action actor-critic (VAAC), is proposed to address the challenge of efficient exploration in RL. This work is inspired by humans' ability to imagine the potential outcomes of their actions without actually taking them. In order to emulate this ability, VAAC introduces a new actor called virtual actor (VA), alongside the conventional actor-critic framework. Unlike the conventional actor, the VA takes the virtual action to anticipate the next state without interacting with the environment. With the virtual policy following a Gaussian distribution, the VA is trained to maximize the anticipated novelty of the subsequent state resulting from a virtual action. If any next state resulting from available actions does not exhibit high anticipated novelty, training the VA leads to an increase in the virtual policy entropy. Hence, high virtual policy entropy represents that there is no room for exploration. The proposed VAAC aims to maximize a modified Q function, which combines cumulative rewards and the negative sum of virtual policy entropy. Experimental results show that the VAAC improves the exploration performance compared to existing algorithms.

Woohyeon Moon, Taeyoung Kim, Bumgeun Park et al.

Transformer is a state-of-the-art model in the field of natural language processing (NLP). Current NLP models primarily increase the number of transformers to improve processing performance. However, this technique requires a lot of training resources such as computing capacity. In this paper, a novel structure of Transformer is proposed. It is featured by full layer normalization, weighted residual connection, positional encoding exploiting reinforcement learning, and zero masked self-attention. The proposed Transformer model, which is called Enhanced Transformer, is validated by the bilingual evaluation understudy (BLEU) score obtained with the Multi30k translation dataset. As a result, the Enhanced Transformer achieves 202.96% higher BLEU score as compared to the original transformer with the translation dataset.

Bumgeun Park, Jihui Lee, Taeyoung Kim et al.

This paper presents a technique to train a robot to perform kick-motion in AI soccer by using reinforcement learning (RL). In RL, an agent interacts with an environment and learns to choose an action in a state at each step. When training RL algorithms, a problem called the curse of dimensionality (COD) can occur if the dimension of the state is high and the number of training data is low. The COD often causes degraded performance of RL models. In the situation of the robot kicking the ball, as the ball approaches the robot, the robot chooses the action based on the information obtained from the soccer field. In order not to suffer COD, the training data, which are experiences in the case of RL, should be collected evenly from all areas of the soccer field over (theoretically infinite) time. In this paper, we attempt to use the relative coordinate system (RCS) as the state for training kick-motion of robot agent, instead of using the absolute coordinate system (ACS). Using the RCS eliminates the necessity for the agent to know all the (state) information of entire soccer field and reduces the dimension of the state that the agent needs to know to perform kick-motion, and consequently alleviates COD. The training based on the RCS is performed with the widely used Deep Q-network (DQN) and tested in the AI Soccer environment implemented with Webots simulation software.

Taeyoung Kim, Dongsoo Har

In multi-goal reinforcement learning with a sparse binary reward, training agents is particularly challenging, due to a lack of successful experiences. To solve this problem, hindsight experience replay (HER) generates successful experiences even from unsuccessful ones. However, generating successful experiences from uniformly sampled ones is not an efficient process. In this paper, the impact of exploiting the property of achieved goals in generating successful experiences is investigated and a novel cluster-based sampling strategy is proposed. The proposed sampling strategy groups episodes with different achieved goals by using a cluster model and samples experiences in the manner of HER to create the training batch. The proposed method is validated by experiments with three robotic control tasks of the OpenAI Gym. The results of experiments demonstrate that the proposed method is substantially sample efficient and achieves better performance than baseline approaches.

Dongyoung Kim, Huiwon Jang, Myungkyu Koo et al.

While Vision-Language-Action models (VLAs) have shown remarkable progress toward human-like generalist robotic policies through the versatile intelligence (i.e. broad scene understanding and language-conditioned generalization) inherited from pre-trained Vision-Language Models, they still struggle with complex real-world tasks requiring broader functional capabilities (e.g. motion awareness, memory-aware decision making, and physical sensing). To address this, we introduce RLDX-1, a general-purpose robotic policy for dexterous manipulation built on the Multi-Stream Action Transformer (MSAT), an architecture that unifies these capabilities by integrating heterogeneous modalities through modality-specific streams with cross-modal joint self-attention. RLDX-1 further combines this architecture with system-level design choices, including synthesizing training data for rare manipulation scenarios, learning procedures specialized for human-like manipulation, and inference optimizations for real-time deployment. Through empirical evaluation, we show that RLDX-1 consistently outperforms recent frontier VLAs (e.g. $π_{0.5}$ and GR00T N1.6) across both simulation benchmarks and real-world tasks that require broad functional capabilities beyond general versatility. In particular, RLDX-1 shows superiority in ALLEX humanoid tasks by achieving success rates of 86.8% while $π_{0.5}$ and GR00T N1.6 achieve around 40%, highlighting the ability of RLDX-1 to control a high-DoF humanoid robot under diverse functional demands. Together, these results position RLDX-1 as a promising step toward reliable VLAs for complex, contact-rich, and dynamic real-world dexterous manipulation.

Taeyoung Kim

We investigate random feature models in which neural networks sampled from a prescribed initialization ensemble are frozen and used as random features, with only the readout weights optimized. Adopting a statistical-physics viewpoint, we study the training, test, and generalization errors beyond the mean-kernel approximation. Since the predictor is a nonlinear functional of the induced random kernel, the ensemble-averaged errors depend not only on the mean kernel but also on higher-order fluctuation statistics. Within an effective field-theoretic framework, these finite-width contributions naturally appear as loop corrections. We derive the loop corrections to the training, test, and generalization errors, obtain their scaling laws, and support the theory with experimental verification.

Taeyoung Kim, Myungjoo Kang

The Rectified Power Unit (RePU) activation function, a differentiable generalization of the Rectified Linear Unit (ReLU), has shown promise in constructing neural networks due to its smoothness properties. However, deep RePU networks often suffer from critical issues such as vanishing or exploding values during training, rendering them unstable regardless of hyperparameter initialization. Leveraging the perspective of effective field theory, we identify the root causes of these failures and propose the Modified Rectified Power Unit (MRePU) activation function. MRePU addresses RePU's limitations while preserving its advantages, such as differentiability and universal approximation properties. Theoretical analysis demonstrates that MRePU satisfies criticality conditions necessary for stable training, placing it in a distinct universality class. Extensive experiments validate the effectiveness of MRePU, showing significant improvements in training stability and performance across various tasks, including polynomial regression, physics-informed neural networks (PINNs) and real-world vision tasks. Our findings highlight the potential of MRePU as a robust alternative for building deep neural networks.

Taeyoung Kim, Myungjoo Kang

Traditionally, classical numerical schemes have been employed to solve partial differential equations (PDEs) using computational methods. Recently, neural network-based methods have emerged. Despite these advancements, neural network-based methods, such as physics-informed neural networks (PINNs) and neural operators, exhibit deficiencies in robustness and generalization. To address these issues, numerous studies have integrated classical numerical frameworks with machine learning techniques, incorporating neural networks into parts of traditional numerical methods. In this study, we focus on hyperbolic conservation laws by replacing traditional numerical fluxes with neural operators. To this end, we developed loss functions inspired by established numerical schemes related to conservation laws and approximated numerical fluxes using Fourier neural operators (FNOs). Our experiments demonstrated that our approach combines the strengths of both traditional numerical schemes and FNOs, outperforming standard FNO methods in several respects. For instance, we demonstrate that our method is robust, has resolution invariance, and is feasible as a data-driven method. In particular, our method can make continuous predictions over time and exhibits superior generalization capabilities with out-of-distribution (OOD) samples, which are challenges that existing neural operator methods encounter.

Wonjae Lee, Taeyoung Kim, Hyungbin Park

This paper introduces an operator-based neural network, the mirror-padded Fourier neural operator (MFNO), designed to learn the dynamics of stochastic systems. MFNO extends the standard Fourier neural operator (FNO) by incorporating mirror padding, enabling it to handle non-periodic inputs. We rigorously prove that MFNOs can approximate solutions of path-dependent stochastic differential equations and Lipschitz transformations of fractional Brownian motions to an arbitrary degree of accuracy. Our theoretical analysis builds on Wong--Zakai type theorems and various approximation techniques. Empirically, the MFNO exhibits strong resolution generalization--a property rarely seen in standard architectures such as LSTMs, TCNs, and DeepONet. Furthermore, our model achieves performance that is comparable or superior to these baselines while offering significantly faster sample path generation than classical numerical schemes.

Taeyoung Kim, Hongseok Yang

The recent theoretical analysis of deep neural networks in their infinite-width limits has deepened our understanding of initialisation, feature learning, and training of those networks, and brought new practical techniques for finding appropriate hyperparameters, learning network weights, and performing inference. In this paper, we broaden this line of research by showing that this infinite-width analysis can be extended to the Jacobian of a deep neural network. We show that a multilayer perceptron (MLP) and its Jacobian at initialisation jointly converge to a Gaussian process (GP) as the widths of the MLP's hidden layers go to infinity and characterise this GP. We also prove that in the infinite-width limit, the evolution of the MLP under the so-called robust training (i.e., training with a regulariser on the Jacobian) is described by a linear first-order ordinary differential equation that is determined by a variant of the Neural Tangent Kernel. We experimentally show the relevance of our theoretical claims to wide finite networks, and empirically analyse the properties of kernel regression solution to obtain an insight into Jacobian regularisation.

Taeyoung Kim, Jimin Lee, Myungkyu Koo et al.

Vision-Language-Action (VLA) models have shown its capabilities in robot manipulation by leveraging rich representations from pre-trained Vision-Language Models (VLMs). However, their representations arguably remain suboptimal, lacking sensitivity to robotic signals such as control actions and proprioceptive states. To address the issue, we introduce Robot State-aware Contrastive Loss (RS-CL), a simple and effective representation regularization for VLA models, designed to bridge the gap between VLM representations and robotic signals. In particular, RS-CL aligns the representations more closely with the robot's proprioceptive states, by using relative distances between the states as soft supervision. Complementing the original action prediction objective, RS-CL effectively enhances control-relevant representation learning, while being lightweight and fully compatible with standard VLA training pipeline. Our empirical results demonstrate that RS-CL substantially improves the manipulation performance of state-of-the-art VLA models; it pushes the prior art from 30.8% to 41.5% on pick-and-place tasks in RoboCasa-Kitchen, through more accurate positioning during grasping and placing, and boosts success rates from 45.0% to 58.3% on challenging real-robot manipulation tasks.

Myungkyu Koo, Daewon Choi, Taeyoung Kim et al.

Inherently, robotic manipulation tasks are history-dependent: leveraging past context could be beneficial. However, most existing Vision-Language-Action models (VLAs) have been designed without considering this aspect, i.e., they rely solely on the current observation, ignoring preceding context. In this paper, we propose HAMLET, a scalable framework to adapt VLAs to attend to the historical context during action prediction. Specifically, we introduce moment tokens that compactly encode perceptual information at each timestep. Their representations are initialized with time-contrastive learning, allowing them to better capture temporally distinctive aspects. Next, we employ a lightweight memory module that integrates the moment tokens across past timesteps into memory features, which are then leveraged for action prediction. Through empirical evaluation, we show that HAMLET successfully transforms a state-of-the-art VLA into a history-aware policy, especially demonstrating significant improvements on long-horizon tasks that require historical context. In particular, on top of GR00T N1.5, HAMLET achieves an average success rate of 76.4% on history-dependent real-world tasks, surpassing the baseline performance by 47.2%. Furthermore, HAMLET pushes prior art performance from 64.1% to 66.4% on RoboCasa Kitchen (100-demo setup) and from 95.6% to 97.7% on LIBERO, highlighting its effectiveness even under generic robot-manipulation benchmarks.

Taeyoung Kim

Fourier Neural Operators (FNOs) have emerged as leading surrogates for solver operators for various functional problems, yet their stability, generalization and frequency behavior lack a principled explanation. We present a systematic effective field theory analysis of FNOs in an infinite dimensional function space, deriving closed recursion relations for the layer kernel and four point vertex and then examining three practically important settings-analytic activations, scale invariant cases and architectures with residual connections. The theory shows that nonlinear activations inevitably couple frequency inputs to high frequency modes that are otherwise discarded by spectral truncation, and experiments confirm this frequency transfer. For wide networks, we derive explicit criticality conditions on the weight initialization ensemble that ensure small input perturbations maintain a uniform scale across depth, and we confirm experimentally that the theoretically predicted ratio of kernel perturbations matches the measurements. Taken together, our results quantify how nonlinearity enables neural operators to capture non-trivial features, supply criteria for hyperparameter selection via criticality analysis, and explain why scale invariant activations and residual connections enhance feature learning in FNOs.

Jaeyoon Kim, Yoonki Cho, Taeyoung Kim et al.

Nearest neighbor (NN) graph based visual re-ranking has emerged as a powerful approach for improving retrieval accuracy, offering the advantages of effectively exploring high-dimensional manifolds without requiring additional fine-tuning. However, the effectiveness of NN graph-based re-ranking is fundamentally constrained by the quality of its edge connectivity, as incorrect connections between dissimilar (negative) images frequently occur. This is known as a noisy edge problem, which hinders the re-ranking performance of existing techniques and limits their potential. To remedy this issue, we propose a complementary denoising method based on Continuous Conditional Random Fields (C-CRF) that leverages statistical distances derived from similarity-based distributions. As a pre-processing step for enhancing NN graph-based retrieval, our approach constructs fully connected cliques around each anchor image and employs a novel statistical distance metric to robustly alleviate noisy edges before re-ranking while achieving efficient processing through offline computation. Extensive experimental results demonstrate that our method consistently improves three different NN graph-based re-ranking approaches, yielding significant gains in retrieval accuracy.

Taeyoung Kim, Youngsoo Ha, Myungjoo Kang

Magnetohydrodynamics (MHD) plays a pivotal role in describing the dynamics of plasma and conductive fluids, essential for understanding phenomena such as the structure and evolution of stars and galaxies, and in nuclear fusion for plasma motion through ideal MHD equations. Solving these hyperbolic PDEs requires sophisticated numerical methods, presenting computational challenges due to complex structures and high costs. Recent advances introduce neural operators like the Fourier Neural Operator (FNO) as surrogate models for traditional numerical analyses. This study explores a modified Flux Fourier neural operator model to approximate the numerical flux of ideal MHD, offering a novel approach that outperforms existing neural operator models by enabling continuous inference, generalization outside sampled distributions, and faster computation compared to classical numerical schemes.

Taeyoung Kim, Luiz Felipe Vecchietti, Kyujin Choi et al.

In multi-agent reinforcement learning, the cooperative learning behavior of agents is very important. In the field of heterogeneous multi-agent reinforcement learning, cooperative behavior among different types of agents in a group is pursued. Learning a joint-action set during centralized training is an attractive way to obtain such cooperative behavior, however, this method brings limited learning performance with heterogeneous agents. To improve the learning performance of heterogeneous agents during centralized training, two-stage heterogeneous centralized training which allows the training of multiple roles of heterogeneous agents is proposed. During training, two training processes are conducted in a series. One of the two stages is to attempt training each agent according to its role, aiming at the maximization of individual role rewards. The other is for training the agents as a whole to make them learn cooperative behaviors while attempting to maximize shared collective rewards, e.g., team rewards. Because these two training processes are conducted in a series in every timestep, agents can learn how to maximize role rewards and team rewards simultaneously. The proposed method is applied to 5 versus 5 AI robot soccer for validation. Simulation results show that the proposed method can train the robots of the robot soccer team effectively, achieving higher role rewards and higher team rewards as compared to other approaches that can be used to solve problems of training cooperative multi-agent.

Taeyoung Kim, Youngsun Kwon, Sung-eui Yoon

This paper proposes a real-time system integrating an acoustic material estimation from visual appearance and an on-the-fly mapping in the 3-dimension. The proposed method estimates the acoustic materials of surroundings in indoor scenes and incorporates them to a 3-D occupancy map, as a robot moves around the environment. To estimate the acoustic material from the visual cue, we apply the state-of-the-art semantic segmentation CNN network based on the assumption that the visual appearance and the acoustic materials have a strong association. Furthermore, we introduce an update policy to handle the material estimations during the online mapping process. As a result, our environment map with acoustic material can be used for sound-related robotics applications, such as sound source localization taking into account various acoustic propagation (e.g., reflection).