Junsung Park

Hyemi Jang, Junsung Park, Dahuin Jung et al.

Although supervised image denoising networks have shown remarkable performance on synthesized noisy images, they often fail in practice due to the difference between real and synthesized noise. Since clean-noisy image pairs from the real world are extremely costly to gather, self-supervised learning, which utilizes noisy input itself as a target, has been studied. To prevent a self-supervised denoising model from learning identical mapping, each output pixel should not be influenced by its corresponding input pixel; This requirement is known as J-invariance. Blind-spot networks (BSNs) have been a prevalent choice to ensure J-invariance in self-supervised image denoising. However, constructing variations of BSNs by injecting additional operations such as downsampling can expose blinded information, thereby violating J-invariance. Consequently, convolutions designed specifically for BSNs have been allowed only, limiting architectural flexibility. To overcome this limitation, we propose PUCA, a novel J-invariant U-Net architecture, for self-supervised denoising. PUCA leverages patch-unshuffle/shuffle to dramatically expand receptive fields while maintaining J-invariance and dilated attention blocks (DABs) for global context incorporation. Experimental results demonstrate that PUCA achieves state-of-the-art performance, outperforming existing methods in self-supervised image denoising.

Junsung Park, Kyungmin Kim, Hyunjung Shim

Existing LiDAR semantic segmentation methods often struggle with performance declines in adverse weather conditions. Previous work has addressed this issue by simulating adverse weather or employing universal data augmentation during training. However, these methods lack a detailed analysis and understanding of how adverse weather negatively affects LiDAR semantic segmentation performance. Motivated by this issue, we identified key factors of adverse weather and conducted a toy experiment to pinpoint the main causes of performance degradation: (1) Geometric perturbation due to refraction caused by fog or droplets in the air and (2) Point drop due to energy absorption and occlusions. Based on these findings, we propose new strategic data augmentation techniques. First, we introduced a Selective Jittering (SJ) that jitters points in the random range of depth (or angle) to mimic geometric perturbation. Additionally, we developed a Learnable Point Drop (LPD) to learn vulnerable erase patterns with a Deep Q-Learning Network to approximate the point drop phenomenon from adverse weather conditions. Without precise weather simulation, these techniques strengthen the LiDAR semantic segmentation model by exposing it to vulnerable conditions identified by our data-centric analysis. Experimental results confirmed the suitability of the proposed data augmentation methods for enhancing robustness against adverse weather conditions. Our method achieves a notable 39.5 mIoU on the SemanticKITTI-to-SemanticSTF benchmark, improving the baseline by 8.1\%p and establishing a new state-of-the-art. Our code will be released at \url{https://github.com/engineerJPark/LiDARWeather}.

Yeongtak Oh, Sangwon Yu, Junsung Park et al.

Despite recent progress in vision-language models (VLMs), existing approaches often fail to generate personalized responses based on the user's specific experiences, as they lack the ability to associate visual inputs with a user's accumulated visual-textual context. We newly formalize this challenge as contextualized visual personalization, which requires the visual recognition and textual retrieval of personalized visual experiences by VLMs when interpreting new images. To address this issue, we propose CoViP, a unified framework that treats personalized image captioning as a core task for contextualized visual personalization and improves this capability through reinforcement-learning-based post-training and caption-augmented generation. We further introduce diagnostic evaluations that explicitly rule out textual shortcut solutions and verify whether VLMs truly leverage visual context. Extensive experiments demonstrate that existing open-source and proprietary VLMs exhibit substantial limitations, while CoViP not only improves personalized image captioning but also yields holistic gains across downstream personalization tasks. These results highlight CoViP as a crucial stage for enabling robust and generalizable contextualized visual personalization.

Anya Zorin, Zilin Si, Myungsun Park et al.

Vision-based learning from demonstrations has achieved remarkable success in enabling robots to perform manipulation tasks and high-level semantic reasoning, yet it remains insufficient for complex, contact-rich manipulation. While there is broad agreement that tactile sensing improves manipulation, there is no empirical guidance on which tactile sensors are best suited for which manipulation tasks. In this paper, we provide a systematic, task-driven evaluation of tactile sensors for robot manipulation and propose a framework for selecting and evaluating sensors based on manipulation policy performance. Separate manipulation policies are trained for tactile sensors of four distinct modalities: visual, acoustic, magnetic, and resistive, across three tasks: pick-and-place with unknown mass, object reorientation, and plug insertion. For each task, an analysis of how sensor properties such as spatial resolution, shear sensing, and tactile representation, and the inherent material friction affect task performances is done. Rather than tactile sensing being universally beneficial in the same way, our results show that the usefulness of tactile information depends strongly on sensor modality, material properties, and the specific manipulation tasks. All of the tactile sensors, code, data, and hardware setup will be publicly available on the project website.

Junsung Park, Hyunjung Shim

Existing Driving VLAs predict trajectories while largely ignoring their visual tokens -- a phenomenon we trace not to insufficient training but to a structurally ill-posed task formulation. We show that trajectory recovery, when viewed through the lens of inverse kinematics, requires both a current and a future visual state as boundary conditions; existing VLAs supply only the former, which encourages the model to shortcut through ego status and text commands alone. To address this, we re-design Driving VLA in the style of an inverse kinematics solver. First, a next visual state prediction objective that requires the LLM to predict the future visual scene provides dense visual supervision and suppresses shortcut paths. Second, a separate Inverse Kinematics Network (a cross-attention-based conditional diffusion model) that takes only the current and future visual states as input is designed to suppress reliance on ego status and textual shortcuts during trajectory decoding. With this simple prescription alone, our 0.5B-scale model recovers visual grounding and reaches trajectory planning performance comparable to 7B--8B VLAs more than an order of magnitude larger, on both the closed-loop NAVSIM-v2 and the nuScenes benchmarks. Extensive analysis further shows that this improvement stems from a recovered ability to exploit visual features, with the effect being most pronounced in dynamic driving situations such as turning.

Junsung Park, Hogun Kee, Songhwai Oh

This paper presents a modality-augmented fine-tuning framework designed to adapt foundation robot policies to diverse humanoid embodiments. We validate our approach across two distinct settings: (i) the GR1 embodiment, utilizing public datasets where we introduce post-processed modalities, including binary contact signals and ZoeDepth-generated metric depth; and (ii) the Unitree G1 embodiment, for which we contribute a novel multi-modal dataset incorporating cuRobo motion planning, inverse kinematics, and ground-truth contact-force measurements. Our experiments demonstrate that modality augmentation consistently enhances policy performance across different embodiments. Specifically, for the GR1, integrating contact-state cues and RGB-D fusion improves online success rates from 51% to 63%. Furthermore, in the G1 "Pick Apple to Bowl" task, our contact-augmented model achieves a success rate of 94%, significantly outperforming the 48% achieved by standard fine-tuning and the 0% baseline of zero-shot transfer. These results highlight that lightweight post-processing effectively strengthens policies for GR1, while high-quality multi-modal data is crucial for reliable transfer to the Unitree G1. Consequently, this work establishes a unified, data-centric pathway for extending foundation robot policies through targeted modality design and multi-modal fine-tuning.

Seungjun Yu, Seonho Lee, Namho Kim et al.

Recent advancements in multimodal large language models (MLLMs) have shown strong understanding of driving scenes, drawing interest in their application to autonomous driving. However, high-level reasoning in safety-critical scenarios, where avoiding one traffic risk can create another, remains a major challenge. Such reasoning is often infeasible with only a single front view and requires a comprehensive view of the environment, which we achieve through multi-view inputs. We define Safety-Critical Reasoning as a new task that leverages multi-view inputs to address this challenge. Then, we distill Safety-Critical Reasoning into two stages: first resolve the immediate risk, then mitigate the decision-induced downstream risks. To support this, we introduce WaymoQA, a dataset of 35,000 human-annotated question-answer pairs covering complex, high-risk driving scenarios. The dataset includes multiple-choice and open-ended formats across both image and video modalities. Experiments reveal that existing MLLMs underperform in safety-critical scenarios compared to normal scenes, but fine-tuning with WaymoQA significantly improves their reasoning ability, highlighting the effectiveness of our dataset in developing safer and more reasoning-capable driving agents. Our code and data are provided in https://github.com/sjyu001/WaymoQA

Saehyung Lee, Sangwon Yu, Junsung Park et al.

In this paper, we primarily address the issue of dialogue-form context query within the interactive text-to-image retrieval task. Our methodology, PlugIR, actively utilizes the general instruction-following capability of LLMs in two ways. First, by reformulating the dialogue-form context, we eliminate the necessity of fine-tuning a retrieval model on existing visual dialogue data, thereby enabling the use of any arbitrary black-box model. Second, we construct the LLM questioner to generate non-redundant questions about the attributes of the target image, based on the information of retrieval candidate images in the current context. This approach mitigates the issues of noisiness and redundancy in the generated questions. Beyond our methodology, we propose a novel evaluation metric, Best log Rank Integral (BRI), for a comprehensive assessment of the interactive retrieval system. PlugIR demonstrates superior performance compared to both zero-shot and fine-tuned baselines in various benchmarks. Additionally, the two methodologies comprising PlugIR can be flexibly applied together or separately in various situations. Our codes are available at https://github.com/Saehyung-Lee/PlugIR.

Jonghyun Lee, Dahuin Jung, Saehyung Lee et al.

Test-time adaptation (TTA) fine-tunes pre-trained deep neural networks for unseen test data. The primary challenge of TTA is limited access to the entire test dataset during online updates, causing error accumulation. To mitigate it, TTA methods have utilized the model output's entropy as a confidence metric that aims to determine which samples have a lower likelihood of causing error. Through experimental studies, however, we observed the unreliability of entropy as a confidence metric for TTA under biased scenarios and theoretically revealed that it stems from the neglect of the influence of latent disentangled factors of data on predictions. Building upon these findings, we introduce a novel TTA method named Destroy Your Object (DeYO), which leverages a newly proposed confidence metric named Pseudo-Label Probability Difference (PLPD). PLPD quantifies the influence of the shape of an object on prediction by measuring the difference between predictions before and after applying an object-destructive transformation. DeYO consists of sample selection and sample weighting, which employ entropy and PLPD concurrently. For robust adaptation, DeYO prioritizes samples that dominantly incorporate shape information when making predictions. Our extensive experiments demonstrate the consistent superiority of DeYO over baseline methods across various scenarios, including biased and wild. Project page is publicly available at https://whitesnowdrop.github.io/DeYO/.

Hyungyu Lee, Saehyung Lee, Hyemi Jang et al.

The disparity in accuracy between classes in standard training is amplified during adversarial training, a phenomenon termed the robust fairness problem. Existing methodologies aimed to enhance robust fairness by sacrificing the model's performance on easier classes in order to improve its performance on harder ones. However, we observe that under adversarial attacks, the majority of the model's predictions for samples from the worst class are biased towards classes similar to the worst class, rather than towards the easy classes. Through theoretical and empirical analysis, we demonstrate that robust fairness deteriorates as the distance between classes decreases. Motivated by these insights, we introduce the Distance-Aware Fair Adversarial training (DAFA) methodology, which addresses robust fairness by taking into account the similarities between classes. Specifically, our method assigns distinct loss weights and adversarial margins to each class and adjusts them to encourage a trade-off in robustness among similar classes. Experimental results across various datasets demonstrate that our method not only maintains average robust accuracy but also significantly improves the worst robust accuracy, indicating a marked improvement in robust fairness compared to existing methods.

Junsung Park, Jungbeom Lee, Jongyoon Song et al.

While CLIP has significantly advanced multimodal understanding by bridging vision and language, the inability to grasp negation - such as failing to differentiate concepts like "parking" from "no parking" - poses substantial challenges. By analyzing the data used in the public CLIP model's pre-training, we posit this limitation stems from a lack of negation-inclusive data. To address this, we introduce data generation pipelines that employ a large language model (LLM) and a multimodal LLM to produce negation-inclusive captions. Fine-tuning CLIP with data generated from our pipelines, we develop NegationCLIP, which enhances negation awareness while preserving the generality. Moreover, to enable a comprehensive evaluation of negation understanding, we propose NegRefCOCOg-a benchmark tailored to test VLMs' ability to interpret negation across diverse expressions and positions within a sentence. Experiments on various CLIP architectures validate the effectiveness of our data generation pipelines in enhancing CLIP's ability to perceive negation accurately. Additionally, NegationCLIP's enhanced negation awareness has practical applications across various multimodal tasks, demonstrated by performance gains in text-to-image generation and referring image segmentation.

Seonho Lee, Jiho Choi, Inha Kang et al.

Vision-Language Models (VLMs) have shown remarkable performance on diverse visual and linguistic tasks, yet they remain fundamentally limited in their understanding of 3D spatial structures. We propose Geometric Distillation, a lightweight, annotation-free fine-tuning framework that injects human-inspired geometric cues into pretrained VLMs without modifying their architecture. By distilling (1) sparse correspondences, (2) relative depth relations, and (3) dense cost volumes from off-the-shelf 3D foundation models (e.g., MASt3R, VGGT), our method shapes representations to be geometry-aware while remaining compatible with natural image-text inputs. Through extensive evaluations on 3D vision-language reasoning and 3D perception benchmarks, our method consistently outperforms prior approaches, achieving improved 3D spatial reasoning with significantly lower computational cost. Our work demonstrates a scalable and efficient path to bridge 2D-trained VLMs with 3D understanding, opening up wider use in spatially grounded multimodal tasks.

Junsung Park, Hwijeong Lee, Inha Kang et al.

Existing domain generalization methods for LiDAR semantic segmentation under adverse weather struggle to accurately predict "things" categories compared to "stuff" categories. In typical driving scenes, "things" categories can be dynamic and associated with higher collision risks, making them crucial for safe navigation and planning. Recognizing the importance of "things" categories, we identify their performance drop as a serious bottleneck in existing approaches. We observed that adverse weather induces degradation of semantic-level features and both corruption of local features, leading to a misprediction of "things" as "stuff". To mitigate these corruptions, we suggest our method, NTN - segmeNt Things for No-accident. To address semantic-level feature corruption, we bind each point feature to its superclass, preventing the misprediction of things classes into visually dissimilar categories. Additionally, to enhance robustness against local corruption caused by adverse weather, we define each LiDAR beam as a local region and propose a regularization term that aligns the clean data with its corrupted counterpart in feature space. NTN achieves state-of-the-art performance with a +2.6 mIoU gain on the SemanticKITTI-to-SemanticSTF benchmark and +7.9 mIoU on the SemanticPOSS-to-SemanticSTF benchmark. Notably, NTN achieves a +4.8 and +7.9 mIoU improvement on "things" classes, respectively, highlighting its effectiveness.

Wonjeong Ryu, Seungjun Yu, Seokha Moon et al.

End-to-End (E2E) planning has become a powerful paradigm for autonomous driving, yet current systems remain fundamentally uncertainty-blind. They assume perception outputs are fully reliable, even in ambiguous or poorly observed scenes, leaving the planner without an explicit measure of uncertainty. To address this limitation, we propose a camera-only E2E framework that estimates aleatoric uncertainty directly in BEV space and incorporates it into planning. Our method produces a dense, uncertainty-aware drivability map that captures both semantic structure and geometric layout at pixel-level resolution. To further promote safe and rule-compliant behavior, we introduce a lane-following regularization that encodes lane structure and traffic norms. This prior stabilizes trajectory planning under normal conditions while preserving the flexibility needed for maneuvers such as overtaking or lane changes. Together, these components enable robust and interpretable trajectory planning, even under challenging uncertainty conditions. Evaluated on the NAVSIM benchmark, our method achieves state-of-the-art performance, delivering substantial gains on both the challenging NAVHARD and NAVSAFE subsets. These results demonstrate that our principled aleatoric uncertainty modeling combined with driving priors significantly advances the safety and reliability of camera-only E2E autonomous driving.

Junsung Park

Offline inverse reinforcement learning (IRL) aims to recover a reward function that explains expert behavior using only fixed demonstration data, without any additional online interaction. We propose BiCQL-ML, a policy-free offline IRL algorithm that jointly optimizes a reward function and a conservative Q-function in a bi-level framework, thereby avoiding explicit policy learning. The method alternates between (i) learning a conservative Q-function via Conservative Q-Learning (CQL) under the current reward, and (ii) updating the reward parameters to maximize the expected Q-values of expert actions while suppressing over-generalization to out-of-distribution actions. This procedure can be viewed as maximum likelihood estimation under a soft value matching principle. We provide theoretical guarantees that BiCQL-ML converges to a reward function under which the expert policy is soft-optimal. Empirically, we show on standard offline RL benchmarks that BiCQL-ML improves both reward recovery and downstream policy performance compared to existing offline IRL baselines.

Seungjun Yu, Junsung Park, Youngsun Lim et al.

We present a two-phase vision-language QA system for autonomous driving that answers high-level perception, prediction, and planning questions. In Phase-1, a large multimodal LLM (Qwen2.5-VL-32B) is conditioned on six-camera inputs, a short temporal window of history, and a chain-of-thought prompt with few-shot exemplars. A self-consistency ensemble (multiple sampled reasoning chains) further improves answer reliability. In Phase-2, we augment the prompt with nuScenes scene metadata (object annotations, ego-vehicle state, etc.) and category-specific question instructions (separate prompts for perception, prediction, planning tasks). In experiments on a driving QA benchmark, our approach significantly outperforms the baseline Qwen2.5 models. For example, using 5 history frames and 10-shot prompting in Phase-1 yields 65.1% overall accuracy (vs.62.61% with zero-shot); applying self-consistency raises this to 66.85%. Phase-2 achieves 67.37% overall. Notably, the system maintains 96% accuracy under severe visual corruption. These results demonstrate that carefully engineered prompts and contextual grounding can greatly enhance high-level driving QA with pretrained vision-language models.

Junsung Park, Hyunjung Shim

Weakly Supervised Semantic Segmentation (WSSS) employs weak supervision, such as image-level labels, to train the segmentation model. Despite the impressive achievement in recent WSSS methods, we identify that introducing weak labels with high mean Intersection of Union (mIoU) does not guarantee high segmentation performance. Existing studies have emphasized the importance of prioritizing precision and reducing noise to improve overall performance. In the same vein, we propose ORANDNet, an advanced ensemble approach tailored for WSSS. ORANDNet combines Class Activation Maps (CAMs) from two different classifiers to increase the precision of pseudo-masks (PMs). To further mitigate small noise in the PMs, we incorporate curriculum learning. This involves training the segmentation model initially with pairs of smaller-sized images and corresponding PMs, gradually transitioning to the original-sized pairs. By combining the original CAMs of ResNet-50 and ViT, we significantly improve the segmentation performance over the single-best model and the naive ensemble model, respectively. We further extend our ensemble method to CAMs from AMN (ResNet-like) and MCTformer (ViT-like) models, achieving performance benefits in advanced WSSS models. It highlights the potential of our ORANDNet as a final add-on module for WSSS models.