Search

Recent experiments and theoretical studies strongly indicate that water exhibits a liquid-liquid phase transition (LLPT) in the supercooled domain. An open question is how the LLPT of water can affect the properties of aqueous solutions. Here, we study the structural and thermodynamic properties of supercooled glycerol-water microdroplets at dilute conditions ($\chi_g=3.2~\%$ glycerol mole fraction). The combination of rapid evaporative cooling with ultrafast X-ray scattering allows us to outrun crystallization and gain access to the deeply supercooled regime down to $T=229.3$ K. We find that the density fluctuations of the glycerol-water solution or, equivalently, its isothermal compressibility, $\kappa_T$, increases upon cooling. This is confirmed by molecular dynamics simulations, which indicate that the presence of glycerol shifts the temperature of maximum $\kappa_T$ from $T=230$ K in pure water down to $T=223$ K in the solution. Our findings elucidate the interplay between the complex behavior of water, including its LLPT, and the properties of aqueous solutions at low temperatures, which can have practical consequences in cryogenic biological applications and cryopreservation techniques.

Multi-hop reasoning, which requires multi-step reasoning based on the supporting documents within a given context, remains challenging for large language models (LLMs). LLMs often struggle to filter out irrelevant documents within the context, and their performance is sensitive to the position of supporting documents within that context. In this paper, we identify an additional challenge: LLMs' performance is also sensitive to the order in which the supporting documents are presented. We refer to this as the misordered context problem. To address this issue, we propose a simple yet effective method called context repetition (CoRe), which involves prompting the model by repeatedly presenting the context to ensure the supporting documents are presented in the optimal order for the model. Using CoRe, we improve the F1 score by up to 30%p on multi-hop QA tasks and increase accuracy by up to 70%p on a synthetic task. Additionally, CoRe helps mitigate the well-known "lost-in-the-middle" problem in LLMs and can be effectively combined with retrieval-based approaches utilizing Chain-of-Thought (CoT) reasoning.

Recent studies have addressed intricate phonological phenomena in French, relying on either extensive linguistic knowledge or a significant amount of sentence-level pronunciation data. However, creating such resources is expensive and non-trivial. To this end, we propose a novel two-step approach that encompasses two pronunciation tasks: grapheme-to-phoneme and post-lexical processing. We then investigate the efficacy of the proposed approach with a notably limited amount of sentence-level pronunciation data. Our findings demonstrate that the proposed two-step approach effectively mitigates the lack of extensive labeled data, and serves as a feasible solution for addressing French phonological phenomena even under resource-constrained environments., Comment: Accepted at EMNLP 2024 Main

Emission lines from complex organic molecules in B335 were observed in four epochs, spanning a luminosity burst of about 10 years duration. The emission lines increased dramatically in intensity as the luminosity increased, but they have decreased only slightly as the luminosity has decreased. This behavior agrees with expectations of rapid sublimation as the dust temperature increases, but slower freeze-out after the dust temperature drops. Further monitoring of this source, along with detailed chemical models, will exploit this natural laboratory for astrochemistry., Comment: Accepted for publication in ApJL

Co-lex partial orders were recently introduced in (Cotumaccio et al., SODA 2021 and JACM 2023) as a powerful tool to index finite state automata, with applications to regular expression matching. They generalize Wheeler orders (Gagie et al., Theoretical Computer Science 2017) and naturally reflect the co-lexicographic order of the strings labeling source-to-node paths in the automaton. Briefly, the co-lex width $p$ of a finite-state automaton measures how sortable its states are with respect to the co-lex order among the strings they accept. Automata of co-lex width $p$ can be compressed to $O(\log p)$ bits per edge and admit regular expression matching algorithms running in time proportional to $p^2$ per matched character. The deterministic co-lex width of a regular language $\mathcal L$ is the smallest width of such a co-lex order, among all DFAs recognizing $\mathcal L$. Since languages of small co-lex width admit efficient solutions to automata compression and pattern matching, computing the co-lex width of a language is relevant in these applications. The paper introducing co-lex orders determined that the deterministic co-lex width $p$ of a language $\mathcal L$ can be computed in time proportional to $m^{O(p)}$, given as input any DFA $\mathcal A$ for $\mathcal L$, of size (number of transitions) $m =|\mathcal A|$. In this paper, using new techniques, we show that it is possible to decide in $O(m^p)$ time if the deterministic co-lex width of the language recognized by a given minimum DFA is strictly smaller than some integer $p\ge 2$. We complement this upper bound with a matching conditional lower bound based on the Strong Exponential Time Hypothesis. The problem is known to be PSPACE-complete when the input is an NFA (D'Agostino et al., Theoretical Computer Science 2023); thus, together with that result, our paper essentially settles the complexity of the problem.

3D dense captioning is a task to localize objects in a 3D scene and generate descriptive sentences for each object. Recent approaches in 3D dense captioning have adopted transformer encoder-decoder frameworks from object detection to build an end-to-end pipeline without hand-crafted components. However, these approaches struggle with contradicting objectives where a single query attention has to simultaneously view both the tightly localized object regions and contextual environment. To overcome this challenge, we introduce SIA (See-It-All), a transformer pipeline that engages in 3D dense captioning with a novel paradigm called late aggregation. SIA simultaneously decodes two sets of queries-context query and instance query. The instance query focuses on localization and object attribute descriptions, while the context query versatilely captures the region-of-interest of relationships between multiple objects or with the global scene, then aggregated afterwards (i.e., late aggregation) via simple distance-based measures. To further enhance the quality of contextualized caption generation, we design a novel aggregator to generate a fully informed caption based on the surrounding context, the global environment, and object instances. Extensive experiments on two of the most widely-used 3D dense captioning datasets demonstrate that our proposed method achieves a significant improvement over prior methods., Comment: Accepted to ACL 2024 Findings

Accurate patient selection is critical in radiotherapy (RT) to prevent ineffective treatments. Traditional survival prediction models, relying on structured data, often lack precision. This study explores the potential of large language models (LLMs) to structure unstructured electronic health record (EHR) data, thereby improving survival prediction accuracy through comprehensive clinical information integration. Data from 34,276 patients treated with RT at Yonsei Cancer Center between 2013 and 2023 were analyzed, encompassing both structured and unstructured data. An open-source LLM was used to structure the unstructured EHR data via single-shot learning, with its performance compared against a domain-specific medical LLM and a smaller variant. Survival prediction models were developed using statistical, machine learning, and deep learning approaches, incorporating both structured and LLM-structured data. Clinical experts evaluated the accuracy of the LLM-structured data. The open-source LLM achieved 87.5% accuracy in structuring unstructured EHR data without additional training, significantly outperforming the domain-specific medical LLM, which reached only 35.8% accuracy. Larger LLMs were more effective, particularly in extracting clinically relevant features like general condition and disease extent, which closely correlated with patient survival. Incorporating LLM-structured clinical features into survival prediction models significantly improved accuracy, with the C-index of deep learning models increasing from 0.737 to 0.820. These models also became more interpretable by emphasizing clinically significant factors. This study shows that general-domain LLMs, even without specific medical training, can effectively structure large-scale unstructured EHR data, substantially enhancing the accuracy and interpretability of clinical predictive models., Comment: 23 pages, 2 tables, 4 figures

In this paper, we introduce a method to tackle Domain Generalized Semantic Segmentation (DGSS) by utilizing domain-invariant semantic knowledge from text embeddings of vision-language models. We employ the text embeddings as object queries within a transformer-based segmentation framework (textual object queries). These queries are regarded as a domain-invariant basis for pixel grouping in DGSS. To leverage the power of textual object queries, we introduce a novel framework named the textual query-driven mask transformer (tqdm). Our tqdm aims to (1) generate textual object queries that maximally encode domain-invariant semantics and (2) enhance the semantic clarity of dense visual features. Additionally, we suggest three regularization losses to improve the efficacy of tqdm by aligning between visual and textual features. By utilizing our method, the model can comprehend inherent semantic information for classes of interest, enabling it to generalize to extreme domains (e.g., sketch style). Our tqdm achieves 68.9 mIoU on GTA5$\rightarrow$Cityscapes, outperforming the prior state-of-the-art method by 2.5 mIoU. The project page is available at https://byeonghyunpak.github.io/tqdm., Comment: ECCV 2024

An index on a finite-state automaton is a data structure able to locate specific patterns on the automaton's paths and consequently on the regular language accepted by the automaton itself. Cotumaccio and Prezza [SODA '21], introduced a data structure able to solve pattern matching queries on automata, generalizing the famous FM-index for strings of Ferragina and Manzini [FOCS '00]. The efficiency of their index depends on the width of a particular partial order of the automaton's states, the smaller the width of the partial order, the faster is the index. However, computing the partial order of minimal width is NP-hard. This problem was mitigated by Cotumaccio [DCC '22], who relaxed the conditions on the partial order, allowing it to be a partial preorder. This relaxation yields the existence of a unique partial preorder of minimal width that can be computed in polynomial time. In the paper at hand, we present a new class of partial preorders and show that they have the following useful properties: (i) they can be computed in polynomial time, (ii) their width is never larger than the width of Cotumaccio's preorders, and (iii) there exist infinite classes of automata on which the width of Cotumaccio's pre-order is linearly larger than the width of our preorder., Comment: 20 pages, 3 figures, submitted in SPIRE 2024

The goal of this work is to simultaneously generate natural talking faces and speech outputs from text. We achieve this by integrating Talking Face Generation (TFG) and Text-to-Speech (TTS) systems into a unified framework. We address the main challenges of each task: (1) generating a range of head poses representative of real-world scenarios, and (2) ensuring voice consistency despite variations in facial motion for the same identity. To tackle these issues, we introduce a motion sampler based on conditional flow matching, which is capable of high-quality motion code generation in an efficient way. Moreover, we introduce a novel conditioning method for the TTS system, which utilises motion-removed features from the TFG model to yield uniform speech outputs. Our extensive experiments demonstrate that our method effectively creates natural-looking talking faces and speech that accurately match the input text. To our knowledge, this is the first effort to build a multimodal synthesis system that can generalise to unseen identities., Comment: CVPR 2024

The LCP array is an important tool in stringology, allowing to speed up pattern matching algorithms and enabling compact representations of the suffix tree. Recently, Conte et al. [DCC 2023] and Cotumaccio et al. [SPIRE 2023] extended the definition of this array to Wheeler DFAs and, ultimately, to arbitrary labeled graphs, proving that it can be used to efficiently solve matching statistics queries on the graph's paths. In this paper, we provide the first efficient algorithm building the LCP array of a directed labeled graph with $n$ nodes and $m$ edges labeled over an alphabet of size $\sigma$. After arguing that the natural generalization of a compact-space LCP-construction algorithm by Beller et al. [J. Discrete Algorithms 2013] runs in time $\Omega(n\sigma)$, we present a new algorithm based on dynamic range stabbing building the LCP array in $O(n\log \sigma)$ time and $O(n\log\sigma)$ bits of working space.

This research explores the integration of language models and unsupervised anomaly detection in medical imaging, addressing two key questions: (1) Can language models enhance the interpretability of anomaly detection maps? and (2) Can anomaly maps improve the generalizability of language models in open-set anomaly detection tasks? To investigate these questions, we introduce a new dataset for multi-image visual question-answering on brain magnetic resonance images encompassing multiple conditions. We propose KQ-Former (Knowledge Querying Transformer), which is designed to optimally align visual and textual information in limited-sample contexts. Our model achieves a 60.81% accuracy on closed questions, covering disease classification and severity across 15 different classes. For open questions, KQ-Former demonstrates a 70% improvement over the baseline with a BLEU-4 score of 0.41, and achieves the highest entailment ratios (up to 71.9%) and lowest contradiction ratios (down to 10.0%) among various natural language inference models. Furthermore, integrating anomaly maps results in an 18% accuracy increase in detecting open-set anomalies, thereby enhancing the language model's generalizability to previously unseen medical conditions. The code and dataset are available at https://github.com/compai-lab/miccai-2024-junli?tab=readme-ov-file, Comment: 13 pages, 7 figures. 5th International Workshop on Multiscale Multimodal Medical Imaging (MMMI 2024)

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., Comment: 44 pages; updated authors list and fixed author names

Employing ab initio many-body perturbation theory combined with dynamical mean field theory, we discovered that in strongly correlated topological semimetals HoPtBi and PrAlGe, which exhibit topological singular points in the vicinity of the Fermi level, the formation of 4$f$ quasiparticles are forbidden. We show that blocking hybridization channels at the topological singular point effectively enhances on-site Coulomb repulsion, resulting in a substantial self-energy. This renders the topological singular point incompatible with the presence of strongly correlated electrons at the Fermi level. In contrast to the Kondo effect, our findings suggest that the topological quasiparticles in close proximity to the singular points do not hybridize with 4$f$ electrons due to the self-energy, thus hindering the manifestation of heavy-fermion behavior when the singular points persist at the Fermi level.

Recently, with increasing interest in pet healthcare, the demand for computer-aided diagnosis (CAD) systems in veterinary medicine has increased. The development of veterinary CAD has stagnated due to a lack of sufficient radiology data. To overcome the challenge, we propose a generative active learning framework based on a variational autoencoder. This approach aims to alleviate the scarcity of reliable data for CAD systems in veterinary medicine. This study utilizes datasets comprising cardiomegaly radiograph data. After removing annotations and standardizing images, we employed a framework for data augmentation, which consists of a data generation phase and a query phase for filtering the generated data. The experimental results revealed that as the data generated through this framework was added to the training data of the generative model, the frechet inception distance consistently decreased from 84.14 to 50.75 on the radiograph. Subsequently, when the generated data were incorporated into the training of the classification model, the false positive of the confusion matrix also improved from 0.16 to 0.66 on the radiograph. The proposed framework has the potential to address the challenges of data scarcity in medical CAD, contributing to its advancement.

An unbiased spectral survey of V883 Ori, an eruptive young star, was carried out with the Atacama Large Millimeter/submillimeter Array (ALMA) in Band 6. The detected line emission from various molecules reveals morphological/kinematical features in both the Keplerian disk and the infalling envelope. A direct infall signature, red-shifted absorption against continuum, has been detected in CO, HCO$^+$, HCN, HNC, and H$_2$CO. HCO$^+$ and SO show large arm-like structures that probably connect from an infalling envelope to the disk. HCN and H$_2$CO reveal a distinct boundary between the inner and outer disk and reveal tentative spiral structures connecting the outer disk to the inner disk. HNC shows a large central emission hole (r $\sim$0.3\arcsec) due to its chemical conversion to HCN at high temperatures. The HDO emission, a direct tracer of the water sublimation region, has been detected in the disk. Molecular emission from complex organic molecules (COMs) is confined within the HDO emission boundary, and HCO$^+$ has an emission hole in its distribution due to its destruction by water. Together, these features suggest that the current episode of eruption in V883 Ori may be triggered by the infall from the envelope to the outer disk, generating a spiral wave, which propagates inward and greatly enhances the accretion onto the central star., Comment: Accepted for publication in ApJ

For object detection task with noisy labels, it is important to consider not only categorization noise, as in image classification, but also localization noise, missing annotations, and bogus bounding boxes. However, previous studies have only addressed certain types of noise (e.g., localization or categorization). In this paper, we propose Universal-Noise Annotation (UNA), a more practical setting that encompasses all types of noise that can occur in object detection, and analyze how UNA affects the performance of the detector. We analyzed the development direction of previous works of detection algorithms and examined the factors that impact the robustness of detection model learning method. We open-source the code for injecting UNA into the dataset and all the training log and weight are also shared., Comment: appendix and code : https://github.com/Ryoo72/UNA

Contrastive Language-Image Pretraining has emerged as a prominent approach for training vision and text encoders with uncurated image-text pairs from the web. To enhance data-efficiency, recent efforts have introduced additional supervision terms that involve random-augmented views of the image. However, since the image augmentation process is unaware of its text counterpart, this procedure could cause various degrees of image-text misalignments during training. Prior methods either disregarded this discrepancy or introduced external models to mitigate the impact of misalignments during training. In contrast, we propose a novel metric learning approach that capitalizes on these misalignments as an additional training source, which we term "Misalign, Contrast then Distill (MCD)". Unlike previous methods that treat augmented images and their text counterparts as simple positive pairs, MCD predicts the continuous scales of misalignment caused by the augmentation. Our extensive experimental results show that our proposed MCD achieves state-of-the-art transferability in multiple classification and retrieval downstream datasets., Comment: ICCV 2023

Recent advances in vision language pretraining (VLP) have been largely attributed to the large-scale data collected from the web. However, uncurated dataset contains weakly correlated image-text pairs, causing data inefficiency. To address the issue, knowledge distillation have been explored at the expense of extra image and text momentum encoders to generate teaching signals for misaligned image-text pairs. In this paper, our goal is to resolve the misalignment problem with an efficient distillation framework. To this end, we propose ECLIPSE: Expediting Contrastive Language-Image Pretraining with Self-distilled Encoders. ECLIPSE features a distinctive distillation architecture wherein a shared text encoder is utilized between an online image encoder and a momentum image encoder. This strategic design choice enables the distillation to operate within a unified projected space of text embedding, resulting in better performance. Based on the unified text embedding space, ECLIPSE compensates for the additional computational cost of the momentum image encoder by expediting the online image encoder. Through our extensive experiments, we validate that there is a sweet spot between expedition and distillation where the partial view from the expedited online image encoder interacts complementarily with the momentum teacher. As a result, ECLIPSE outperforms its counterparts while achieving substantial acceleration in inference speed., Comment: AAAI 2024

Protostellar outflows often present a knotty appearance, providing evidence of sporadic accretion in stellar mass growth. To understand the direct relation between mass accretion and ejection, we analyze the contemporaneous accretion activity and associated ejection components in B335. B335 has brightened in the mid-IR by 2.5 mag since 2010, indicating increased luminosity, presumably due to increased mass accretion rate onto the protostar. ALMA observations of 12CO emission in the outflow reveal high-velocity emission, estimated to have been ejected 4.6 - 2 years before the ALMA observation and consistent with the jump in mid-IR brightness. The consistency in timing suggests that the detected high-velocity ejection components are directly linked to the most recent accretion activity. We calculated the kinetic energy, momentum, and force for the ejection component associated with the most recent accretion activity and found that at least, about 1.0% of accreted mass has been ejected. More accurate information on the jet inclination and the temperature of the ejected gas components will better constrain the ejected mass induced by the recently enhanced accretion event., Comment: 11 pages, 5 figures, and 1 table