arXiv subjects

Jungwon Park

Publications and source records attributed to Jungwon Park.

At least 19 recordsLinked to original sources

arXiv2026

Orthogonal Negative Guidance in Attention Feature Space for Text-to-Image Generation

Text-to-image (T2I) models have become increasingly capable of generating high-quality images. Yet, enforcing the explicit absence of a specified object or attribute remains a fundamentally challenging problem. Existing approaches, including prompt negation, post-hoc editing, and negative guidance, remain insufficient for explicit concept suppression, often failing to remove the target concept or degrading overall image quality. To this end, we propose Orthogonal Negative Guidance in attention feature space, a training-free method that operates in the attention output space of MM-DiT-based T2I transformers. Our method orthogonalizes negative-prompt attention features with respect to positive-prompt features and subtracts only the orthogonal component, suppressing unwanted concepts while preserving desired semantics. Experiments on FLUX-dev and FLUX-schnell show that our method achieves favorable trade-offs between concept suppression, prompt alignment, and image quality. In human evaluation, our method outperforms the second-best baseline by 18.78%. We further show that our method supports multi-concept suppression and adjustable concept suppression.

cs.CV↗

arXiv2026

When Confidence Misleads: Suffix Anchoring and Anchor-Proximity Confidence Modulation for Diffusion Language Models

Diffusion language models generate text by iteratively selecting and denoising masked positions, making position selection a central inference-time decision. Most training-free methods rely on model confidence, assuming that high-confidence positions are ready to be decoded. However, this assumption can fail in fully non-autoregressive (fully non-AR) decoding, where selection spans the entire masked response sequence. This can cause incomplete generation, partly due to EOT overconfidence. Existing remedies directly suppress EOT tokens, but provide limited gains or require additional fine-tuning. We show that placing even a semantically minimal suffix anchor near the end of the response region discourages premature EOT prediction and promotes complete generation. However, despite providing little supporting context, the anchor makes nearby positions highly confident and causes them to be decoded too early. We therefore propose Suffix-Anchored Confidence Modulation, a simple training-free method that uses suffix anchoring with progress-dependent confidence modulation near the anchor. Across text-only reasoning, vision-language reasoning, and code-generation benchmarks, our method outperforms diverse fully non-AR and semi-AR baselines. It also achieves particularly large gains under highly parallel decoding, demonstrating the benefit of improving the fully non-AR regime over block-wise semi-AR regime.

cs.CL↗

arXiv2026

Selective Aggregation of Attention Maps Improves Diffusion-Based Visual Interpretation

Numerous studies on text-to-image (T2I) generative models have utilized cross-attention maps to boost application performance and interpret model behavior. However, the distinct characteristics of attention maps from different attention heads remain relatively underexplored. In this study, we show that selectively aggregating cross-attention maps from heads most relevant to a target concept can improve visual interpretability. Compared to the diffusion-based segmentation method DAAM, our approach achieves higher mean IoU scores. We also find that the most relevant heads capture concept-specific features more accurately than the least relevant ones, and that selective aggregation helps diagnose prompt misinterpretations. These findings suggest that attention head selection offers a promising direction for improving the interpretability and controllability of T2I generation.

cs.CV↗

arXiv2025

DOS: Directional Object Separation in Text Embeddings for Multi-Object Image Generation

Recent progress in text-to-image (T2I) generative models has led to significant improvements in generating high-quality images aligned with text prompts. However, these models still struggle with prompts involving multiple objects, often resulting in object neglect or object mixing. Through extensive studies, we identify four problematic scenarios, Similar Shapes, Similar Textures, Dissimilar Background Biases, and Many Objects, where inter-object relationships frequently lead to such failures. Motivated by two key observations about CLIP embeddings, we propose DOS (Directional Object Separation), a method that modifies three types of CLIP text embeddings before passing them into text-to-image models. Experimental results show that DOS consistently improves the success rate of multi-object image generation and reduces object mixing. In human evaluations, DOS significantly outperforms four competing methods, receiving 26.24%-43.04% more votes across four benchmarks. These results highlight DOS as a practical and effective solution for improving multi-object image generation.

cs.CV↗

arXiv2025

Soft Head Selection for Injecting ICL-Derived Task Embeddings

Large language models (LLMs) are commonly adapted to downstream tasks using parameter-efficient fine-tuning (PEFT) or in-context learning (ICL). Recently, ICL-driven embedding-based adaptation has been proposed as a distinct task adaptation paradigm. It derives task-specific embeddings from intermediate activations using few-shot prompts and injects them during inference. Despite its conceptual appeal, this approach has not demonstrated consistent performance gains over PEFT or ICL, and its empirical advantages have been limited in practice. We propose Soft head-selection for ICL-derived Task Embeddings (SITE), a gradient-based method that identifies task-relevant attention heads to enable effective task embedding injection. Across various types of open-ended generation, reasoning, and natural language understanding tasks, SITE significantly outperforms prior embedding-based adaptation methods and few-shot ICL, while using substantially fewer trainable parameters than PEFT. Experiments on 12 LLMs ranging from 4B to 70B parameters demonstrate the generality of our approach, and intra-task and inter-task activation patching analyses further provide new mechanistic insights by revealing strong task dependence in attention head functionality.

cs.CL↗

arXiv2025

ReFlex: Text-Guided Editing of Real Images in Rectified Flow via Mid-Step Feature Extraction and Attention Adaptation

Rectified Flow text-to-image models surpass diffusion models in image quality and text alignment, but adapting ReFlow for real-image editing remains challenging. We propose a new real-image editing method for ReFlow by analyzing the intermediate representations of multimodal transformer blocks and identifying three key features. To extract these features from real images with sufficient structural preservation, we leverage mid-step latent, which is inverted only up to the mid-step. We then adapt attention during injection to improve editability and enhance alignment to the target text. Our method is training-free, requires no user-provided mask, and can be applied even without a source prompt. Extensive experiments on two benchmarks with nine baselines demonstrate its superior performance over prior methods, further validated by human evaluations confirming a strong user preference for our approach.

cs.CV↗

arXiv2025

MC-Swarm: Minimal-Communication Multi-Agent Trajectory Planning and Deadlock Resolution for Quadrotor Swarm

For effective multi-agent trajectory planning, it is important to consider lightweight communication and its potential asynchrony. This paper presents a distributed trajectory planning algorithm for a quadrotor swarm that operates asynchronously and requires no communication except during the initial planning phase. Moreover, our algorithm guarantees no deadlock under asynchronous updates and absence of communication during flight. To effectively ensure these points, we build two main modules: coordination state updater and trajectory optimizer. The coordination state updater computes waypoints for each agent toward its goal and performs subgoal optimization while considering deadlocks, as well as safety constraints with respect to neighbor agents and obstacles. Then, the trajectory optimizer generates a trajectory that ensures collision avoidance even with the asynchronous planning updates of neighboring agents. We provide a theoretical guarantee of collision avoidance with deadlock resolution and evaluate the effectiveness of our method in complex simulation environments, including random forests and narrow-gap mazes. Additionally, to reduce the total mission time, we design a faster coordination state update using lightweight communication. Lastly, our approach is validated through extensive simulations and real-world experiments with cluttered environment scenarios.

cs.RO↗

arXiv2024

Task-Specific Preconditioner for Cross-Domain Few-Shot Learning

Cross-Domain Few-Shot Learning~(CDFSL) methods typically parameterize models with task-agnostic and task-specific parameters. To adapt task-specific parameters, recent approaches have utilized fixed optimization strategies, despite their potential sub-optimality across varying domains or target tasks. To address this issue, we propose a novel adaptation mechanism called Task-Specific Preconditioned gradient descent~(TSP). Our method first meta-learns Domain-Specific Preconditioners~(DSPs) that capture the characteristics of each meta-training domain, which are then linearly combined using task-coefficients to form the Task-Specific Preconditioner. The preconditioner is applied to gradient descent, making the optimization adaptive to the target task. We constrain our preconditioners to be positive definite, guiding the preconditioned gradient toward the direction of steepest descent. Empirical evaluations on the Meta-Dataset show that TSP achieves state-of-the-art performance across diverse experimental scenarios.

cs.LG↗

arXiv2024

Cross-Attention Head Position Patterns Can Align with Human Visual Concepts in Text-to-Image Generative Models

Recent text-to-image diffusion models leverage cross-attention layers, which have been effectively utilized to enhance a range of visual generative tasks. However, our understanding of cross-attention layers remains somewhat limited. In this study, we introduce a mechanistic interpretability approach for diffusion models by constructing Head Relevance Vectors (HRVs) that align with human-specified visual concepts. An HRV for a given visual concept has a length equal to the total number of cross-attention heads, with each element representing the importance of the corresponding head for the given visual concept. To validate HRVs as interpretable features, we develop an ordered weakening analysis that demonstrates their effectiveness. Furthermore, we propose concept strengthening and concept adjusting methods and apply them to enhance three visual generative tasks. Our results show that HRVs can reduce misinterpretations of polysemous words in image generation, successfully modify five challenging attributes in image editing, and mitigate catastrophic neglect in multi-concept generation. Overall, our work provides an advancement in understanding cross-attention layers and introduces new approaches for fine-controlling these layers at the head level.

cs.CV↗

arXiv2024

DMVC-Tracker: Distributed Multi-Agent Trajectory Planning for Target Tracking Using Dynamic Buffered Voronoi and Inter-Visibility Cells

This letter presents a distributed trajectory planning method for multi-agent aerial tracking. The proposed method uses a Dynamic Buffered Voronoi Cell (DBVC) and a Dynamic Inter-Visibility Cell (DIVC) to formulate the distributed trajectory generation. Specifically, the DBVC and the DIVC are time-variant spaces that prevent mutual collisions and occlusions among agents, while enabling them to maintain suitable distances from the moving target. We combine the DBVC and the DIVC with an efficient Bernstein polynomial motion primitive-based tracking generation method, which has been refined into a less conservative approach than in our previous work. The proposed algorithm can compute each agent's trajectory within several milliseconds on an Intel i7 desktop. We validate the tracking performance in challenging scenarios, including environments with dozens of obstacles.

cs.RO↗

arXiv2024

BEACON -- Automated Aberration Correction for Scanning Transmission Electron Microscopy using Bayesian Optimization

Aberration correction is an important aspect of modern high-resolution scanning transmission electron microscopy. Most methods of aligning aberration correctors require specialized sample regions and are unsuitable for fine-tuning aberrations without interrupting on-going experiments. Here, we present an automated method of correcting first- and second-order aberrations called BEACON which uses Bayesian optimization of the normalized image variance to efficiently determine the optimal corrector settings. We demonstrate its use on gold nanoparticles and a hafnium dioxide thin film showing its versatility in nano- and atomic-scale experiments. BEACON can correct all first- and second-order aberrations simultaneously to achieve an initial alignment and first- and second-order aberrations independently for fine alignment. Ptychographic reconstructions are used to demonstrate an improvement in probe shape and a reduction in the target aberration.

cond-mat.mtrl-sci↗

arXiv2024

BPMP-Tracker: A Versatile Aerial Target Tracker Using Bernstein Polynomial Motion Primitives

This letter presents a versatile trajectory planning pipeline for aerial tracking. The proposed tracker is capable of handling various chasing settings such as complex unstructured environments, crowded dynamic obstacles and multiple-target following. Among the entire pipeline, we focus on developing a predictor for future target motion and a chasing trajectory planner. For rapid computation, we employ the sample-check-select strategy: modules sample a set of candidate movements, check multiple constraints, and then select the best trajectory. Also, we leverage the properties of Bernstein polynomials for quick calculations. The prediction module predicts the trajectories of the targets, which do not overlap with static and dynamic obstacles. Then the trajectory planner outputs a trajectory, ensuring various conditions such as occlusion and collision avoidance, the visibility of all targets within a camera image and dynamical limits. We fully test the proposed tracker in simulations and hardware experiments under challenging scenarios, including dual-target following, environments with dozens of dynamic obstacles and complex indoor and outdoor spaces.

cs.RO↗

arXiv2024

Selectively Informative Description can Reduce Undesired Embedding Entanglements in Text-to-Image Personalization

In text-to-image personalization, a timely and crucial challenge is the tendency of generated images overfitting to the biases present in the reference images. We initiate our study with a comprehensive categorization of the biases into background, nearby-object, tied-object, substance (in style re-contextualization), and pose biases. These biases manifest in the generated images due to their entanglement into the subject embedding. This undesired embedding entanglement not only results in the reflection of biases from the reference images into the generated images but also notably diminishes the alignment of the generated images with the given generation prompt. To address this challenge, we propose SID~(Selectively Informative Description), a text description strategy that deviates from the prevalent approach of only characterizing the subject's class identification. SID is generated utilizing multimodal GPT-4 and can be seamlessly integrated into optimization-based models. We present comprehensive experimental results along with analyses of cross-attention maps, subject-alignment, non-subject-disentanglement, and text-alignment.

cs.CV↗

arXiv2023

Isotropic Representation Can Improve Dense Retrieval

The recent advancement in language representation modeling has broadly affected the design of dense retrieval models. In particular, many of the high-performing dense retrieval models evaluate representations of query and document using BERT, and subsequently apply a cosine-similarity based scoring to determine the relevance. BERT representations, however, are known to follow an anisotropic distribution of a narrow cone shape and such an anisotropic distribution can be undesirable for the cosine-similarity based scoring. In this work, we first show that BERT-based DR also follows an anisotropic distribution. To cope with the problem, we introduce unsupervised post-processing methods of Normalizing Flow and whitening, and develop token-wise method in addition to the sequence-wise method for applying the post-processing methods to the representations of dense retrieval models. We show that the proposed methods can effectively enhance the representations to be isotropic, then we perform experiments with ColBERT and RepBERT to show that the performance (NDCG at 10) of document re-ranking can be improved by 5.17\%$\sim$8.09\% for ColBERT and 6.88\%$\sim$22.81\% for RepBERT. To examine the potential of isotropic representation for improving the robustness of DR models, we investigate out-of-distribution tasks where the test dataset differs from the training dataset. The results show that isotropic representation can achieve a generally improved performance. For instance, when training dataset is MS-MARCO and test dataset is Robust04, isotropy post-processing can improve the baseline performance by up to 24.98\%. Furthermore, we show that an isotropic model trained with an out-of-distribution dataset can even outperform a baseline model trained with the in-distribution dataset.

cs.CL↗

arXiv2023

Decentralized Deadlock-free Trajectory Planning for Quadrotor Swarm in Obstacle-rich Environments -- Extended version

This paper presents a decentralized multi-agent trajectory planning (MATP) algorithm that guarantees to generate a safe, deadlock-free trajectory in an obstacle-rich environment under a limited communication range. The proposed algorithm utilizes a grid-based multi-agent path planning (MAPP) algorithm for deadlock resolution, and we introduce the subgoal optimization method to make the agent converge to the waypoint generated from the MAPP without deadlock. In addition, the proposed algorithm ensures the feasibility of the optimization problem and collision avoidance by adopting a linear safe corridor (LSC). We verify that the proposed algorithm does not cause a deadlock in both random forests and dense mazes regardless of communication range, and it outperforms our previous work in flight time and distance. We validate the proposed algorithm through a hardware demonstration with ten quadrotors.

cs.RO↗

arXiv2023

QP Chaser: Polynomial Trajectory Generation for Autonomous Aerial Tracking

Maintaining the visibility of the target is one of the major objectives of aerial tracking missions. This paper proposes a target-visible trajectory planning pipeline using quadratic programming. Our approach can handle various tracking settings, including single and dual target following and both static and dynamic environments, unlike other works that focus on a single specific setup. In contrast to other studies that fully trust the predicted trajectory of the target and consider only the visibility of the center of the target, our pipeline considers error in target path prediction and the entire body of the target to maintain the target visibility robustly. First, a prediction module uses a sample-check strategy to quickly calculate the reachable areas of moving objects, which represent the areas their bodies can reach, considering obstacles. Subsequently, the planning module formulates a single QP problem, considering path homotopy, to generate a tracking trajectory that maximizes the visibility of the target's reachable area among obstacles. The performance of the planner is validated in multiple scenarios, through high-fidelity simulations and real-world experiments.

cs.RO↗

arXiv2022

Multiplication of freestanding semiconductor membranes from a single wafer by advanced remote epitaxy

Freestanding single-crystalline membranes are an important building block for functional electronics. Especially, compounds semiconductor membranes such as III-N and III-V offer great opportunities for optoelectronics, high-power electronics, and high-speed computing. Despite huge efforts to produce such membranes by detaching epitaxial layers from donor wafers, however, it is still challenging to harvest epitaxial layers using practical processes. Here, we demonstrate a method to grow and harvest multiple epitaxial membranes with extremely high throughput at the wafer scale. For this, 2D materials are directly formed on III-N and III-V substrates in epitaxy systems, which enables an advanced remote epitaxy scheme comprised of multiple alternating layers of 2D materials and epitaxial layers that can be formed by a single epitaxy run. Each epilayer in the multi-stack structure is then harvested by layer-by-layer peeling, producing multiple freestanding membranes with unprecedented throughput from a single wafer. Because 2D materials allow peeling at the interface without damaging the epilayer or the substrate, wafers can be reused for subsequent membrane production. Therefore, this work represents a meaningful step toward high-throughput and low-cost production of single-crystal membranes that can be heterointegrated.

physics.app-ph↗

arXiv2022

Online Distributed Trajectory Planning for Quadrotor Swarm with Feasibility Guarantee using Linear Safe Corridor

This paper presents a new online multi-agent trajectory planning algorithm that guarantees to generate safe, dynamically feasible trajectories in a cluttered environment. The proposed algorithm utilizes a linear safe corridor (LSC) to formulate the distributed trajectory optimization problem with only feasible constraints, so it does not resort to slack variables or soft constraints to avoid optimization failure. We adopt a priority-based goal planning method to prevent the deadlock without an additional procedure to decide which robot to yield. The proposed algorithm can compute the trajectories for 60 agents on average 15.5 ms per agent with an Intel i7 laptop and shows a similar flight distance and distance compared to the baselines based on soft constraints. We verified that the proposed method can reach the goal without deadlock in both the random forest and the indoor space, and we validated the safety and operability of the proposed algorithm through a real flight test with ten quadrotors in a maze-like environment.

cs.RO↗