Exploring Multimodal Perception in Large Language Models Through Perceptual Strength Ratings

This study investigated the multimodal perception of large language models (LLMs), focusing on their ability to capture human-like perceptual strength ratings across sensory modalities. Utilizing perceptual strength ratings as a benchmark, the research compared GPT-3.5, GPT-4, GPT-4o, and GPT-4o-mini, highlighting the influence of multimodal inputs on grounding and linguistic reasoning. While GPT-4 and GPT-4o demonstrated strong alignment with human evaluations and significant advancements over smaller models, qualitative analyses revealed distinct differences in processing patterns, such as multisensory overrating and reliance on loose semantic associations. Despite integrating multimodal capabilities, GPT-4o did not exhibit superior grounding compared to GPT-4, raising questions about their role in improving human-like grounding. These findings underscore how LLMs' reliance on linguistic patterns can both approximate and diverge from human embodied cognition, revealing limitations in replicating sensory experiences.

A Neural Operator-Based Emulator for Regional Shallow Water Dynamics

Coastal regions are particularly vulnerable to the impacts of rising sea levels and extreme weather events. Accurate real-time forecasting of hydrodynamic processes in these areas is essential for infrastructure planning and climate adaptation. In this study, we present the Multiple-Input Temporal Operator Network (MITONet), a novel autoregressive neural emulator that employs dimensionality reduction to efficiently approximate high-dimensional numerical solvers for complex, nonlinear problems that are governed by time-dependent, parameterized partial differential equations. Although MITONet is applicable to a wide range of problems, we showcase its capabilities by forecasting regional tide-driven dynamics described by the two-dimensional shallow-water equations, while incorporating initial conditions, boundary conditions, and a varying domain parameter. We demonstrate MITONet's performance in a real-world application, highlighting its ability to make accurate predictions by extrapolating both in time and parametric space.

Multi-fidelity Hamiltonian Monte Carlo

Numerous applications in biology, statistics, science, and engineering require generating samples from high-dimensional probability distributions. In recent years, the Hamiltonian Monte Carlo (HMC) method has emerged as a state-of-the-art Markov chain Monte Carlo technique, exploiting the shape of such high-dimensional target distributions to efficiently generate samples. Despite its impressive empirical success and increasing popularity, its wide-scale adoption remains limited due to the high computational cost of gradient calculation. Moreover, applying this method is impossible when the gradient of the posterior cannot be computed (for example, with black-box simulators). To overcome these challenges, we propose a novel two-stage Hamiltonian Monte Carlo algorithm with a surrogate model. In this multi-fidelity algorithm, the acceptance probability is computed in the first stage via a standard HMC proposal using an inexpensive differentiable surrogate model, and if the proposal is accepted, the posterior is evaluated in the second stage using the high-fidelity (HF) numerical solver. Splitting the standard HMC algorithm into these two stages allows for approximating the gradient of the posterior efficiently, while producing accurate posterior samples by using HF numerical solvers in the second stage. We demonstrate the effectiveness of this algorithm for a range of problems, including linear and nonlinear Bayesian inverse problems with in-silico data and experimental data. The proposed algorithm is shown to seamlessly integrate with various low-fidelity and HF models, priors, and datasets. Remarkably, our proposed method outperforms the traditional HMC algorithm in both computational and statistical efficiency by several orders of magnitude, all while retaining or improving the accuracy in computed posterior statistics.

HyperCLOVA X Technical Report

We introduce HyperCLOVA X, a family of large language models (LLMs) tailored to the Korean language and culture, along with competitive capabilities in English, math, and coding. HyperCLOVA X was trained on a balanced mix of Korean, English, and code data, followed by instruction-tuning with high-quality human-annotated datasets while abiding by strict safety guidelines reflecting our commitment to responsible AI. The model is evaluated across various benchmarks, including comprehensive reasoning, knowledge, commonsense, factuality, coding, math, chatting, instruction-following, and harmlessness, in both Korean and English. HyperCLOVA X exhibits strong reasoning capabilities in Korean backed by a deep understanding of the language and cultural nuances. Further analysis of the inherent bilingual nature and its extension to multilingualism highlights the model's cross-lingual proficiency and strong generalization ability to untargeted languages, including machine translation between several language pairs and cross-lingual inference tasks. We believe that HyperCLOVA X can provide helpful guidance for regions or countries in developing their sovereign LLMs.

Efficient Diffusion-Driven Corruption Editor for Test-Time Adaptation

Test-time adaptation (TTA) addresses the unforeseen distribution shifts occurring during test time. In TTA, both performance and, memory and time consumption serve as crucial considerations. A recent diffusion-based TTA approach for restoring corrupted images involves image-level updates. However, using pixel space diffusion significantly increases resource requirements compared to conventional model updating TTA approaches, revealing limitations as a TTA method. To address this, we propose a novel TTA method by leveraging a latent diffusion model (LDM) based image editing model and fine-tuning it with our newly introduced corruption modeling scheme. This scheme enhances the robustness of the diffusion model against distribution shifts by creating (clean, corrupted) image pairs and fine-tuning the model to edit corrupted images into clean ones. Moreover, we introduce a distilled variant to accelerate the model for corruption editing using only 4 network function evaluations (NFEs). We extensively validated our method across various architectures and datasets including image and video domains. Our model achieves the best performance with a 100 times faster runtime than that of a diffusion-based baseline. Furthermore, it outpaces the speed of the model updating TTA method based on data augmentation threefold, rendering an image-level updating approach more practical.

SF$^2$: Source-free Domain Adaptation Through the Lens of Data Augmentation

In the face of the deep learning model's vulnerability to domain shift, source-free domain adaptation (SFDA) methods have been proposed to adapt models to new, unseen target domains without requiring access to source domain data. Although the potential benefits of applying data augmentation to SFDA are attractive, several challenges arise such as the dependence on prior knowledge of class-preserving transformations and the increase in memory and computational requirements. In this paper, we propose Source-free Domain Adaptation Through the Lens of Data Augmentation (SF(DA)$^2$), a novel approach that leverages the benefits of data augmentation without suffering from these challenges. We construct an augmentation graph in the feature space of the pretrained model using the neighbor relationships between target features and propose spectral neighborhood clustering to identify partitions in the prediction space. Furthermore, we propose implicit feature augmentation and feature disentanglement as regularization loss functions that effectively utilize class semantic information within the feature space. These regularizers simulate the inclusion of an unlimited number of augmented target features into the augmentation graph while minimizing computational and memory demands. Our method shows superior adaptation performance in SFDA scenarios, including 2D image and 3D point cloud datasets and a highly imbalanced dataset.

Entropy is not Enough for Test-Time Adaptation: From the Perspective of Disentangled Factors

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/.

One-Shot Structure-Aware Stylized Image Synthesis

While GAN-based models have been successful in image stylization tasks, they often struggle with structure preservation while stylizing a wide range of input images. Recently, diffusion models have been adopted for image stylization but still lack the capability to maintain the original quality of input images. Building on this, we propose OSASIS: a novel one-shot stylization method that is robust in structure preservation. We show that OSASIS is able to effectively disentangle the semantics from the structure of an image, allowing it to control the level of content and style implemented to a given input. We apply OSASIS to various experimental settings, including stylization with out-of-domain reference images and stylization with text-driven manipulation. Results show that OSASIS outperforms other stylization methods, especially for input images that were rarely encountered during training, providing a promising solution to stylization via diffusion models.

Gradient Alignment with Prototype Feature for Fully Test-time Adaptation

In context of Test-time Adaptation(TTA), we propose a regularizer, dubbed Gradient Alignment with Prototype feature (GAP), which alleviates the inappropriate guidance from entropy minimization loss from misclassified pseudo label. We developed a gradient alignment loss to precisely manage the adaptation process, ensuring that changes made for some data don't negatively impact the model's performance on other data. We introduce a prototype feature of a class as a proxy measure of the negative impact. To make GAP regularizer feasible under the TTA constraints, where model can only access test data without labels, we tailored its formula in two ways: approximating prototype features with weight vectors of the classifier, calculating gradient without back-propagation. We demonstrate GAP significantly improves TTA methods across various datasets, which proves its versatility and effectiveness.

Noise Map Guidance: Inversion with Spatial Context for Real Image Editing

Text-guided diffusion models have become a popular tool in image synthesis, known for producing high-quality and diverse images. However, their application to editing real images often encounters hurdles primarily due to the text condition deteriorating the reconstruction quality and subsequently affecting editing fidelity. Null-text Inversion (NTI) has made strides in this area, but it fails to capture spatial context and requires computationally intensive per-timestep optimization. Addressing these challenges, we present Noise Map Guidance (NMG), an inversion method rich in a spatial context, tailored for real-image editing. Significantly, NMG achieves this without necessitating optimization, yet preserves the editing quality. Our empirical investigations highlight NMG's adaptability across various editing techniques and its robustness to variants of DDIM inversions.