Search

Massive online user review platforms, with their star ratings and text reviews, are reshaping the information available for consumer and public service decisions. We study the leading K-12 schooling platform, GreatSchools, applying machine learning (Natural Language Processing, NLP) to 600,000 reviews that encompass the vast majority of the nation's traditional public, charter, and private schools (84,009 schools in total), supplemented with qualitative analysis of a subsample of reviews. Encompassing more than fifty million words of text, our initial analysis pre-specified eight broad topics and 27 sub-topics and coded review words into these categories. We find that parents write the vast majority of reviews and tend to write more about School Staff and School Culture than students. More generally, text reviews vary in important ways across user types (parents, students, teachers, principals), school sector (traditional public, charter, and private schools), grade level, and demographics. The partial correlation between topics and star ratings also differs across user types and sectors. Taken together, these results suggest that user reviews are less useful than they appear and less useful than with other kinds of products. Our analysis points to design features that might improve their usefulness. The variation in content and value of the reviews also has methodological implications as it shows how NLP can complement qualitative research methods with such large volumes of text.

This paper examines the effects of Pell Grant eligibility on student outcomes. Using a regression discontinuity (RD) design and a partial identification approach, the study provides bounds on the treatment effects that account for selection bias arising from the loss of grant eligibility. While initial eligibility is determined by financial need alone, students must achieve Satisfactory Academic Progress (SAP) to retain the grant. Students eligible for the maximum grant aid are 26 percentage points less likely to persist in the year they lose grant eligibility than those with less aid. This negative effect on persistence extends to graduation; these students are 8 percentage points less likely to graduate within 4 years. Furthermore, these two groups of students differ in their underlying characteristics, which introduces attrition bias into the estimates. Finally, to address this selection bias, I provide bounds on the effects of Pell Grant on student outcomes. While naive RD estimates find no effect on completion rates, bounding estimates reveal that students eligible for the maximum grant aid are up to 4, 2, and 2 percentage points more likely to graduate from a 4-year institution within 4, 5, and 6 years compared to those with less aid, respectively. Furthermore, these eligible students graduate with a higher GPA than previously estimated. These positive effects are larger than those found in earlier studies.

Families are being given more and more choices of schools, from charter schools to virtual schools and private schools, through school vouchers. However, families often report not being well informed about these choices. Many report being awash in information, though not necessarily the information they want or need. State, district, and school websites, along with Facebook pages and word-of-mouth, provide additional information but not necessarily a coherent picture of the various schooling options available to families. Massive school information and review platforms add another source of information. These platforms provide access to much of the same data found on other public websites--test scores, student demographics, graduation rates, and more--combined with text reviews and star ratings provided by students, parents, educators, and other stakeholders. The authors studied the content and usefulness of user reviews from one of the most widely used platforms, GreatSchools, including 50 million words of text across more than 600,000 reviews written about 84,000 schools from 2009 to 2019. The findings suggest that user reviews have the potential to inform consumers about school choices. However, they face many hurdles, some of which have not been well recognized. The authors propose ways to improve school user reviews and discuss implications for school choice.

Over the past decade, there has been a significant increase in the number of U.S. institutions offering STEM-eligible degree programs in economics. This paper documents the trends in STEM-degree offerings across degree levels and examines the share of foreign students and other characteristics of institutions that offer STEM-eligible programs. Using a difference-in-differences design, this paper finds that departments with a proportion of foreign students above the sample median are 6 and 9 percentage points more likely to offer a STEM-eligible degree program at the bachelor's and master's levels, respectively, after the STEM designation in 2013. Additionally, the Tobit regression results suggest that early adopters of STEM-eligible programs are associated with a higher share of foreign students, private institutions and doctoral and research institutions.

Facial expression recognition (FER) remains a challenging task due to label ambiguity caused by the subjective nature of facial expressions and noisy samples. Additionally, class imbalance, which is common in real-world datasets, further complicates FER. Although many studies have shown impressive improvements, they typically address only one of these issues, leading to suboptimal results. To tackle both challenges simultaneously, we propose a novel framework called Navigating Label Ambiguity (NLA), which is robust under real-world conditions. The motivation behind NLA is that dynamically estimating and emphasizing ambiguous samples at each iteration helps mitigate noise and class imbalance by reducing the model's bias toward majority classes. To achieve this, NLA consists of two main components: Noise-aware Adaptive Weighting (NAW) and consistency regularization. Specifically, NAW adaptively assigns higher importance to ambiguous samples and lower importance to noisy ones, based on the correlation between the intermediate prediction scores for the ground truth and the nearest negative. Moreover, we incorporate a regularization term to ensure consistent latent distributions. Consequently, NLA enables the model to progressively focus on more challenging ambiguous samples, which primarily belong to the minority class, in the later stages of training. Extensive experiments demonstrate that NLA outperforms existing methods in both overall and mean accuracy, confirming its robustness against noise and class imbalance. To the best of our knowledge, this is the first framework to address both problems simultaneously., Comment: Accepted by AAAI2025

Using data samples of 102 million Upsilon(1S) events and 158 million Upsilon(2S) events collected by the Belle detector at the KEKB asymmetric-energy e+e- collider, we search for [udsccbar] pentaquark states decaying to Jpsi Lambda. Using the first observations of Upsilon(1S, 2S) inclusive decays to Jpsi Lambda, we find evidence of the P_ccbars(4459)0 state with a significance of 3.3 standard deviations, including statistical and systematic uncertainties. We measure the mass and width of the Pccbars(4459)0 to be (4471.7 +- 4.8 +- 0.6) MeV/c2 and (21.9 +- 13.1 +- 2.7) MeV, respectively. The branching fractions for P_ccbars(4459)0 production are measured to be B[Upsilon(1S) -> P_ccbars(4459)0/ Pbar_ccbars(4459)0 + anything] = (3.5 +- 2.0 +- 0.2)*10-6 and B[Upsilin(2S) -> P_ccbars(4459)0/ Pbar_ccbars(4459)0 +anything] = (2.9 +- 1.7 +- 0.4)*10-6. The inclusive branching fractions of Upsilon(1S, 2S) -> Jpsi Lambda/Lambdabar are measured to be B[Upsilin(1S) -> Jpsi Lambda/Lambdabar + anything] = (36.9 +- 5.3 +- 2.4)*10-6 and B[Upsilon(2S) -> Jpsi Lambda/Lambdabar + anything] = (22.3 +- 5.7 +- 3.1)*10-6. We measure the visible cross section sigma(e+e- -> Jpsi Lambda/Lambdabar + anything) = (90 +- 14 +- 6) fb for the continuum production at sqrt(s) = 10.52 GeV. In all cases, the first uncertainties are statistical and the second are systematic., Comment: 8 pages, 3 figures

Low-rank adaptation (LoRA) has become a standard approach for fine-tuning large foundation models. However, our theoretical understanding of LoRA remains limited as prior analyses of LoRA's training dynamics either rely on linearization arguments or consider highly simplified setups. In this work, we analyze the LoRA loss landscape without such restrictive assumptions. We define two regimes: a ``special regime'', which includes idealized setups where linearization arguments hold, and a ``generic regime'' representing more realistic setups where linearization arguments do not hold. In the generic regime, we show that LoRA training converges to a global minimizer with low rank and small magnitude, or a qualitatively distinct solution with high rank and large magnitude. Finally, we argue that the zero-initialization and weight decay in LoRA training induce an implicit bias toward the low-rank, small-magnitude region of the parameter space -- where global minima lie -- thus shedding light on why LoRA training usually succeeds in finding global minima.

Ion trap system is a leading candidate for quantum network privileged by its long coherence time, high-fidelity gate operations, and the ion-photon entanglement that generates an ideal pair of a stationary memory qubit and a flying communication qubit. Rapid developments in nonlinear quantum frequency conversion techniques have enhanced the potential for constructing a trapped ion quantum network via optical fiber connections. The generation of long-distance entanglement has been demonstrated with ions such as Ca$^{+}$ and Ba$^{+}$, which emit photons in visible or near-infrared range naturally. On the other hand, as the qubit-native photons reside in ultra-violet (UV) spectrum, the Yb$^{+}$ ion has not been considered as a strong competitor for telecommunication qubits despite extensive research on it. Here, we demonstrate an efficient difference-frequency conversion of UV photons, emitted from a trapped Yb$^{+}$ ion, into a visible range. We provide experimental evidence that confirms the converted photons are radiated from the Yb$^{+}$ ion. Our results provide a crucial step toward realizing a long-distance trapped ion quantum network based on Yb$^{+}$ ions through quantum frequency conversion., Comment: 12 pages, 6 figures, 2 tables.; The following article has been accepted by Applied Physics Letters

With the success of transformer architectures across diverse applications, the error correction code transformer (ECCT) has gained significant attention for its superior decoding performance. In spite of its advantages, the error floor phenomenon in ECCT decoding remains unexplored. We present the first investigation of the error floor issue in ECCT and propose a hybrid decoding approach that integrates hard decision decoders as pre- and post-decoders with ECCT to effectively lower the error floor. In particular, we introduce a novel loss function for ECCT that considers the dynamics of hybrid decoding algorithm. Training ECCT with the proposed loss function enhances its ability to correct specific error patterns by taking into account its interaction with the auxiliary decoders. Simulation results demonstrate that the proposed hybrid decoder with the novel loss function significantly outperforms the original ECCT in both the waterfall and the error floor regions., Comment: 6 pages

In the era of artificial intelligence (AI), Transformer demonstrates its performance across various applications. The excessive amount of parameters incurs high latency and energy overhead when processed in the von Neumann architecture. Processing-in-memory (PIM) has shown the potential in accelerating data-intensive applications by reducing data movement. While previous works mainly optimize the computational part of PIM to enhance energy efficiency, the importance of memory design, which consumes the most power in PIM, has been rather neglected. In this work, we present RED, an energy optimization framework for eDRAM-based PIM. We first analyze the PIM operations in eDRAM, obtaining two key observations: 1) memory access energy consumption is predominant in PIM, and 2) read bitline (RBL) voltage swing, sense amplifier power, and retention time are in trade-off relations. Leveraging them, we propose a novel reconfigurable eDRAM and retention-aware scheduling that minimizes the runtime energy consumption of the eDRAM macro. The framework pinpoints the optimal operating point by pre-estimating energy consumption across all possible tiling schemes and memory operations. Then, the reconfigurable eDRAM controls the RBL voltage swing at runtime according to the scheduling, optimizing the memory access power. Moreover, RED employs refresh skipping and sense amplifier power gating to mitigate the energy consumption overhead coming from the trade-off relation. Finally, the RED framework achieves up to 3.05x higher energy efficiency than the prior SRAM-based PIM, reducing the energy consumption of eDRAM macro up to 74.88% with reconfigurable eDRAM and optimization schemes, requiring only 3.5% area and 0.77% energy overhead for scheduling., Comment: 10 pages, 14 figures

We establish the asymptotic stability of solutions to the inflow problem for the one-dimensional barotropic Navier-Stokes equations in half space. When the boundary value is located at the subsonic regime, all the possible thirteen asymptotic patterns are classified in \cite{M01}. We consider the most complicated pattern, the superposition of the boundary layer solution, the 1-rarefaction wave, and the viscous 2-shock waves. In this superposition, the boundary layer is degenerate and large. We prove that, if the strengths of the rarefaction wave and shock wave are small, and if the initial data is a small perturbation of the superposition, then the solution asymptotically converges to the superposition up to a dynamical shift for the shock. As a corollary, our result implies the asymptotic stability for the simpler case where the superposition consists of the degenerate boundary layer solution and the viscous 2-shock. Therefore, we complete the study of the asymptotic stability of the inflow problem for the 1D barotropic Navier-Stokes equations for subsonic boundary values.

Existing methods for simulating individual identities often oversimplify human complexity, which may lead to incomplete or flattened representations. To address this, we introduce SPeCtrum, a grounded framework for constructing authentic LLM agent personas by incorporating an individual's multidimensional self-concept. SPeCtrum integrates three core components: Social Identity (S), Personal Identity (P), and Personal Life Context (C), each contributing distinct yet interconnected aspects of identity. To evaluate SPeCtrum's effectiveness in identity representation, we conducted automated and human evaluations. Automated evaluations using popular drama characters showed that Personal Life Context (C)-derived from short essays on preferences and daily routines-modeled characters' identities more effectively than Social Identity (S) and Personal Identity (P) alone and performed comparably to the full SPC combination. In contrast, human evaluations involving real-world individuals found that the full SPC combination provided a more comprehensive self-concept representation than C alone. Our findings suggest that while C alone may suffice for basic identity simulation, integrating S, P, and C enhances the authenticity and accuracy of real-world identity representation. Overall, SPeCtrum offers a structured approach for simulating individuals in LLM agents, enabling more personalized human-AI interactions and improving the realism of simulation-based behavioral studies., Comment: 21 pages, 8 figures, 5 tables, Accepted in NAACL2025 Main

Evaluating writing quality is complex and time-consuming often delaying feedback to learners. While automated writing evaluation tools are effective for English, Korean automated writing evaluation tools face challenges due to their inability to address multi-view analysis, error propagation, and evaluation explainability. To overcome these challenges, we introduce UKTA (Unified Korean Text Analyzer), a comprehensive Korea text analysis and writing evaluation system. UKTA provides accurate low-level morpheme analysis, key lexical features for mid-level explainability, and transparent high-level rubric-based writing scores. Our approach enhances accuracy and quadratic weighted kappa over existing baseline, positioning UKTA as a leading multi-perspective tool for Korean text analysis and writing evaluation., Comment: Accepted by SAC 2025

There is growing concern over the safety of powerful diffusion models (DMs), as they are often misused to produce inappropriate, not-safe-for-work (NSFW) content or generate copyrighted material or data of individuals who wish to be forgotten. Many existing methods tackle these issues by heavily relying on text-based negative prompts or extensively retraining DMs to eliminate certain features or samples. In this paper, we take a radically different approach, directly modifying the sampling trajectory by leveraging a negation set (e.g., unsafe images, copyrighted data, or datapoints needed to be excluded) to avoid specific regions of data distribution, without needing to retrain or fine-tune DMs. We formally derive the relationship between the expected denoised samples that are safe and those that are not safe, leading to our $\textit{safe}$ denoiser which ensures its final samples are away from the area to be negated. Inspired by the derivation, we develop a practical algorithm that successfully produces high-quality samples while avoiding negation areas of the data distribution in text-conditional, class-conditional, and unconditional image generation scenarios. These results hint at the great potential of our training-free safe denoiser for using DMs more safely., Comment: Preprint

Compute-in-memory (CIM) is an efficient method for implementing deep neural networks (DNNs) but suffers from substantial overhead from analog-to-digital converters (ADCs), especially as ADC precision increases. Low-precision ADCs can re- duce this overhead but introduce partial-sum quantization errors degrading accuracy. Additionally, low-bit weight constraints, im- posed by cell limitations and the need for multiple cells for higher- bit weights, present further challenges. While fine-grained partial- sum quantization has been studied to lower ADC resolution effectively, weight granularity, which limits overall partial-sum quantized accuracy, remains underexplored. This work addresses these challenges by aligning weight and partial-sum quantization granularities at the column-wise level. Our method improves accuracy while maintaining dequantization overhead, simplifies training by removing two-stage processes, and ensures robustness to memory cell variations via independent column-wise scale factors. We also propose an open-source CIM-oriented convolution framework to handle fine-grained weights and partial-sums effi- ciently, incorporating a novel tiling method and group convolution. Experimental results on ResNet-20 (CIFAR-10, CIFAR-100) and ResNet-18 (ImageNet) show accuracy improvements of 0.99%, 2.69%, and 1.01%, respectively, compared to the best-performing related works. Additionally, variation analysis reveals the robust- ness of our method against memory cell variations. These findings highlight the effectiveness of our quantization scheme in enhancing accuracy and robustness while maintaining hardware efficiency in CIM-based DNN implementations. Our code is available at https://github.com/jiyoonkm/ColumnQuant.

Multivariate (MTV) porous materials exhibit unique structural complexities based on their diverse spatial arrangements of multiple building block combinations. These materials possess potential synergistic functionalities that exceed the sum of their individual components. However, the exponentially increasing design complexity of these materials poses significant challenges for accurate ground-state configuration prediction and design. To address this, we propose a Hamiltonian model for quantum computing that integrates compositional, structural, and balance constraints directly into the Hamiltonian, enabling efficient optimization of the MTV configurations. The model employs a graph-based representation to encode linker types as qubits. Our framework leads to exponentially efficient exploration of a vast search space of the linkers to identify optimal configurations based on predefined design variables. To validate our model, a variational quantum circuit was constructed and executed using the Sampling VQE algorithm in the IBM Qiskit. Simulations on experimentally known MTV porous materials (e.g. Cu-THQ-HHTP, Py-MVDBA-COF, MUF-7, and SIOC-COF2) successfully reproduced their ground-state configurations, demonstrating the validity of our model., Comment: 43 pages, 6 figures, 3 supplementary figures, 3 supplementary tables

This paper introduces a novel six-axis force/torque sensor tailored for compact and lightweight legged robots. Unlike traditional strain gauge-based sensors, the proposed non-contact design employs photocouplers, enhancing resistance to physical impacts and reducing damage risk. This approach simplifies manufacturing, lowers costs, and meets the demands of legged robots by combining small size, light weight, and a wide force measurement range. A methodology for optimizing sensor parameters is also presented, focusing on maximizing sensitivity and minimizing error. Precise modeling and analysis of objective functions enabled the derivation of optimal design parameters. The sensor's performance was validated through extensive testing and integration into quadruped robots, demonstrating alignment with theoretical modeling. The sensor's precise measurement capabilities make it suitable for diverse robotic environments, particularly in analyzing interactions between robot feet and the ground. This innovation addresses existing sensor limitations while contributing to advancements in robotics and sensor technology, paving the way for future applications in robotic systems., Comment: 14 pages, 12 figures

Group decision-making often benefits from diverse perspectives, yet power imbalances and social influence can stifle minority opinions and compromise outcomes. This prequel introduces an AI-mediated communication system that leverages the Large Language Model to serve as a devil's advocate, representing underrepresented viewpoints without exposing minority members' identities. Rooted in persuasive communication strategies and anonymity, the system aims to improve psychological safety and foster more inclusive decision-making. Our multi-agent architecture, which consists of a summary agent, conversation agent, AI duplicate checker, and paraphrase agent, encourages the group's critical thinking while reducing repetitive outputs. We acknowledge that reliance on text-based communication and fixed intervention timings may limit adaptability, indicating pathways for refinement. By focusing on the representation of minority viewpoints anonymously in power-imbalanced settings, this approach highlights how AI-driven methods can evolve to support more divergent and inclusive group decision-making., Comment: 10 pages, 3 figures. This is a preprint version of the poster accepted to ACM Conference on Intelligent User Interfaces(IUI'25)

Process Reward Models (PRMs) have proven effective at enhancing mathematical reasoning for Large Language Models (LLMs) by leveraging increased inference-time computation. However, they are predominantly trained on mathematical data and their generalizability to non-mathematical domains has not been rigorously studied. In response, this work first shows that current PRMs have poor performance in other domains. To address this limitation, we introduce VersaPRM, a multi-domain PRM trained on synthetic reasoning data generated using our novel data generation and annotation method. VersaPRM achieves consistent performance gains across diverse domains. For instance, in the MMLU-Pro category of Law, VersaPRM via weighted majority voting, achieves a 7.9% performance gain over the majority voting baseline -- surpassing Qwen2.5-Math-PRM's gain of 1.3%. We further contribute to the community by open-sourcing all data, code and models for VersaPRM.

The Pandora Software Development Kit and algorithm libraries perform reconstruction of neutrino interactions in liquid argon time projection chamber detectors. Pandora is the primary event reconstruction software used at the Deep Underground Neutrino Experiment, which will operate four large-scale liquid argon time projection chambers at the far detector site in South Dakota, producing high-resolution images of charged particles emerging from neutrino interactions. While these high-resolution images provide excellent opportunities for physics, the complex topologies require sophisticated pattern recognition capabilities to interpret signals from the detectors as physically meaningful objects that form the inputs to physics analyses. A critical component is the identification of the neutrino interaction vertex. Subsequent reconstruction algorithms use this location to identify the individual primary particles and ensure they each result in a separate reconstructed particle. A new vertex-finding procedure described in this article integrates a U-ResNet neural network performing hit-level classification into the multi-algorithm approach used by Pandora to identify the neutrino interaction vertex. The machine learning solution is seamlessly integrated into a chain of pattern-recognition algorithms. The technique substantially outperforms the previous BDT-based solution, with a more than 20\% increase in the efficiency of sub-1\,cm vertex reconstruction across all neutrino flavours., Comment: 32 pages, 18 figures

Transparent object manipulation remains a sig- nificant challenge in robotics due to the difficulty of acquiring accurate and dense depth measurements. Conventional depth sensors often fail with transparent objects, resulting in in- complete or erroneous depth data. Existing depth completion methods struggle with interframe consistency and incorrectly model transparent objects as Lambertian surfaces, leading to poor depth reconstruction. To address these challenges, we propose TranSplat, a surface embedding-guided 3D Gaussian Splatting method tailored for transparent objects. TranSplat uses a latent diffusion model to generate surface embeddings that provide consistent and continuous representations, making it robust to changes in viewpoint and lighting. By integrating these surface embeddings with input RGB images, TranSplat effectively captures the complexities of transparent surfaces, enhancing the splatting of 3D Gaussians and improving depth completion. Evaluations on synthetic and real-world transpar- ent object benchmarks, as well as robot grasping tasks, show that TranSplat achieves accurate and dense depth completion, demonstrating its effectiveness in practical applications. We open-source synthetic dataset and model: https://github. com/jeongyun0609/TranSplat, Comment: 7 pages, 6 figures

Previous research has revealed the potential of large language models (LLMs) to support cognitive reframing therapy; however, their focus was primarily on text-based methods, often overlooking the importance of non-verbal evidence crucial in real-life therapy. To alleviate this gap, we extend the textual cognitive reframing to multimodality, incorporating visual clues. Specifically, we present a new dataset called Multi Modal-Cognitive Support Conversation (M2CoSC), which pairs each GPT-4-generated dialogue with an image that reflects the virtual client's facial expressions. To better mirror real psychotherapy, where facial expressions lead to interpreting implicit emotional evidence, we propose a multi-hop psychotherapeutic reasoning approach that explicitly identifies and incorporates subtle evidence. Our comprehensive experiments with both LLMs and vision-language models (VLMs) demonstrate that the VLMs' performance as psychotherapists is significantly improved with the M2CoSC dataset. Furthermore, the multi-hop psychotherapeutic reasoning method enables VLMs to provide more thoughtful and empathetic suggestions, outperforming standard prompting methods., Comment: NAACL 2025 Main

Paired inspiratory-expiratory CT scans enable the quantification of gas trapping due to small airway disease and emphysema by analyzing lung tissue motion in COPD patients. Deformable image registration of these scans assesses regional lung volumetric changes. However, variations in reconstruction kernels between paired scans introduce errors in quantitative analysis. This work proposes a two-stage pipeline to harmonize reconstruction kernels and perform deformable image registration using data acquired from the COPDGene study. We use a cycle generative adversarial network (GAN) to harmonize inspiratory scans reconstructed with a hard kernel (BONE) to match expiratory scans reconstructed with a soft kernel (STANDARD). We then deformably register the expiratory scans to inspiratory scans. We validate harmonization by measuring emphysema using a publicly available segmentation algorithm before and after harmonization. Results show harmonization significantly reduces emphysema measurement inconsistencies, decreasing median emphysema scores from 10.479% to 3.039%, with a reference median score of 1.305% from the STANDARD kernel as the target. Registration accuracy is evaluated via Dice overlap between emphysema regions on inspiratory, expiratory, and deformed images. The Dice coefficient between inspiratory emphysema masks and deformably registered emphysema masks increases significantly across registration stages (p<0.001). Additionally, we demonstrate that deformable registration is robust to kernel variations., Comment: Accepted at SPIE Medical Imaging 2025, Clinical and Biomedical Imaging

We present a measurement of the branching fraction of $B^+\to\tau^+\nu_\tau$ decays using $(387\pm6)\times 10^6$ $\Upsilon(4S)$ collected between 2019 and 2022 with the Belle II detector at the SuperKEKB $e^+e^-$ collider. We reconstruct the accompanying $B^-$ meson using the hadronic tagging method, while $B^+\to\tau^+\nu_\tau$ candidates are identified in the recoil. We find evidence for $B^+\to\tau^+\nu_\tau$ decays at 3.0 standard deviations, including systematic uncertainties. The measured branching fraction is $\mathcal{B}(B^+\to\tau^+\nu_\tau) = [1.24 \pm 0.41 (\text{stat.}) \pm 0.19 (\text{syst.})] \times 10^{-4}$.

The rapid evolution of Large Language Models (LLMs) has enabled the industry to develop various AI-based services. Instruction tuning is considered essential in adapting foundation models for target domains to provide high-quality services to customers. A key challenge in instruction tuning is obtaining high-quality instruction data. Self-Instruct, which automatically generates instruction data using ChatGPT APIs, alleviates the data scarcity problem. To improve the quality of instruction data, Self-Instruct discards many of the instructions generated from ChatGPT, even though it is inefficient in terms of cost owing to many useless API calls. To generate high-quality instruction data at a low cost, we propose a novel data generation framework, Self-Direct Instruction generation (SeDi-Instruct), which employs diversity-based filtering and iterative feedback task generation. Diversity-based filtering maintains model accuracy without excessively discarding low-quality generated instructions by enhancing the diversity of instructions in a batch. This reduces the cost of synthesizing instruction data. The iterative feedback task generation integrates instruction generation and training tasks and utilizes information obtained during the training to create high-quality instruction sets. Our results show that SeDi-Instruct enhances the accuracy of AI models by 5.2%, compared with traditional methods, while reducing data generation costs by 36%., Comment: 12 pages, 12 figures

Importance sampling, which involves sampling from a probability density function (PDF) proportional to the product of an importance weight function and a base PDF, is a powerful technique with applications in variance reduction, biased or customized sampling, data augmentation, and beyond. Inspired by the growing availability of score-based generative models (SGMs), we propose an entirely training-free Importance sampling framework that relies solely on an SGM for the base PDF. Our key innovation is realizing the importance sampling process as a backward diffusion process, expressed in terms of the score function of the base PDF and the specified importance weight function--both readily available--eliminating the need for any additional training. We conduct a thorough analysis demonstrating the method's scalability and effectiveness across diverse datasets and tasks, including importance sampling for industrial and natural images with neural importance weight functions. The training-free aspect of our method is particularly compelling in real-world scenarios where a single base distribution underlies multiple biased sampling tasks, each requiring a different importance weight function. To the best of our knowledge our approach is the first importance sampling framework to achieve this., Comment: 18 pages

We propose and experimentally demonstrate an energy-efficient approach for holding and rearranging an N x M atom array using only N optical tweezers. This is achieved through the sequential release and recapture of M single atoms by a single optical tweezer. By employing a stroboscopic harmonic potential, the phase-space quadrature of the atom's probability distribution can be maintained under this "blinking" potential, provided the trap frequency meets the appropriate conditions. Proof-of-principle experiments confirm that a blinking tweezer can trap M atoms while requiring only 1 / M of the power per atom, and it can even facilitate rearrangement, demonstrated with arrays of up to M = 9 atoms. This method offers a scalable and reconfigurable platform for optical tweezer arrays, crucial for the preparation and manipulation of large-scale qubit systems., Comment: 11 pages, 4 figures

We investigate the quasiparticles of a single nodal ring semimetal SrAs$_3$ through axis-resolved magneto-optical measurements. We observe three types of Landau levels scaling as $\varepsilon \sim \sqrt{B}$, $\varepsilon \sim B^{2/3}$, and $\varepsilon \sim B$ that correspond to Dirac, semi-Dirac, and classical fermions, respectively. Through theoretical analysis, we identify the distinct origins of these three types of fermions present within the nodal ring. In particular, semi-Dirac fermions--a novel type of fermion that can give rise to a range of unique quantum phenomena--emerge from the endpoints of the nodal ring where the energy band disperses linearly along one direction and quadratically along the perpendicular direction, a feature not achievable in nodal point or line structures. The capacity of the nodal ring to simultaneously host multiple fermion types, including semi-Dirac fermions, establishes it as a valuable platform to expand the understanding of topological semimetals., Comment: 35 pages, 23 figures

Autistic students often face challenges in social interaction, which can hinder their educational and personal development. This study introduces Echo-Teddy, a Large Language Model (LLM)-based social robot designed to support autistic students in developing social and communication skills. Unlike previous chatbot-based solutions, Echo-Teddy leverages advanced LLM capabilities to provide more natural and adaptive interactions. The research addresses two key questions: (1) What are the design principles and initial prototype characteristics of an effective LLM-based social robot for autistic students? (2) What improvements can be made based on developer reflection-on-action and expert interviews? The study employed a mixed-methods approach, combining prototype development with qualitative analysis of developer reflections and expert interviews. Key design principles identified include customizability, ethical considerations, and age-appropriate interactions. The initial prototype, built on a Raspberry Pi platform, features custom speech components and basic motor functions. Evaluation of the prototype revealed potential improvements in areas such as user interface, educational value, and practical implementation in educational settings. This research contributes to the growing field of AI-assisted special education by demonstrating the potential of LLM-based social robots in supporting autistic students. The findings provide valuable insights for future developments in accessible and effective social support tools for special education.

The CALICE technological RPC-based SDHCAL prototype that fullfils all the requirements of compactness, hermeticity and power budget of the future lepton accelerator experiments, has been extensively tested and has provided excellent results in terms of the energy resolution and shower separation. A new phase of R&D to validate completely the SDHCAL option for the International Linear Detector (ILD) project of the ILC and also the Circular Electron Positron Collider (CEPC and FCCee) has started with the conception and the realization of new prototypes. The new prototype proposes to exploit the excellent time resolution that can be provided by multi-gap resistive plate chamber detectors in order to better build the hadronic showers with the aim to separate close-by ones and also to single out the contribution of delayed neutrons with the purpose to improve on the Particle Flow Algorithm performances and better reconstruct the showers energy. A new technique to build multi-gap resistive plate chambers has been developed and first results confirm the excellent efficiency of the new detectors. The timing performance is under study using the PETIROC ASIC developed by the OMEGA group. The progress realized on the different aspects of the new concept will be presented and the future steps will be discussed., Comment: 13 pages, 6 figures, Proceedings of the XVII Conference on Resistive Plate Chambers and Related Detectors, 9-13 Sept. 2024, Santiago de Compostela, Spain

This study introduces a motion-based learning network with a global-local self-attention module (MoGLo-Net) to enhance 3D reconstruction in handheld photoacoustic and ultrasound (PAUS) imaging. Standard PAUS imaging is often limited by a narrow field of view and the inability to effectively visualize complex 3D structures. The 3D freehand technique, which aligns sequential 2D images for 3D reconstruction, faces significant challenges in accurate motion estimation without relying on external positional sensors. MoGLo-Net addresses these limitations through an innovative adaptation of the self-attention mechanism, which effectively exploits the critical regions, such as fully-developed speckle area or high-echogenic tissue area within successive ultrasound images to accurately estimate motion parameters. This facilitates the extraction of intricate features from individual frames. Additionally, we designed a patch-wise correlation operation to generate a correlation volume that is highly correlated with the scanning motion. A custom loss function was also developed to ensure robust learning with minimized bias, leveraging the characteristics of the motion parameters. Experimental evaluations demonstrated that MoGLo-Net surpasses current state-of-the-art methods in both quantitative and qualitative performance metrics. Furthermore, we expanded the application of 3D reconstruction technology beyond simple B-mode ultrasound volumes to incorporate Doppler ultrasound and photoacoustic imaging, enabling 3D visualization of vasculature. The source code for this study is publicly available at: https://github.com/guhong3648/US3D

The outstanding opto-electronic properties of lead halide perovskites have been related to the formation of polarons. Nevertheless, the observation of the atomistic deformation brought about by one electron-hole pair in these materials has remained elusive. Here, we measure the diffraction patterns of single CsPbBr$_3$ quantum dots (QDs) with and without resonant excitation in the single exciton limit using serial femtosecond crystallography (SFX). By reconstructing the 3D differential diffraction pattern, we observe small shifts of the Bragg peaks indicative of a crystal-wide deformation field. Building on DFT calculations, we show that these shifts are consistent with the lattice distortion induced by a delocalized electron and a localized hole, forming a mixed large/small exciton polaron. This result creates a clear picture of the polaronic deformation in CsPbBr$_3$ QDs, highlights the exceptional sensitivity of SFX to lattice distortions in few-nanometer crystallites, and establishes an experimental platform for future studies of electron-lattice interactions., Comment: Main: 12 pages, 5 figures; Supplemental: 21 pages, 11 figures

This paper introduces a learning-based visual planner for agile drone flight in cluttered environments. The proposed planner generates collision-free waypoints in milliseconds, enabling drones to perform agile maneuvers in complex environments without building separate perception, mapping, and planning modules. Learning-based methods, such as behavior cloning (BC) and reinforcement learning (RL), demonstrate promising performance in visual navigation but still face inherent limitations. BC is susceptible to compounding errors due to limited expert imitation, while RL struggles with reward function design and sample inefficiency. To address these limitations, this paper proposes an inverse reinforcement learning (IRL)-based framework for high-speed visual navigation. By leveraging IRL, it is possible to reduce the number of interactions with simulation environments and improve capability to deal with high-dimensional spaces while preserving the robustness of RL policies. A motion primitive-based path planning algorithm collects an expert dataset with privileged map data from diverse environments, ensuring comprehensive scenario coverage. By leveraging both the acquired expert and learner dataset gathered from the agent's interactions with the simulation environments, a robust reward function and policy are learned across diverse states. While the proposed method is trained in a simulation environment only, it can be directly applied to real-world scenarios without additional training or tuning. The performance of the proposed method is validated in both simulation and real-world environments, including forests and various structures. The trained policy achieves an average speed of 7 m/s and a maximum speed of 8.8 m/s in real flight experiments. To the best of our knowledge, this is the first work to successfully apply an IRL framework for high-speed visual navigation of drones., Comment: 18 pages, 11 figures, 58 references, and appendix is included

Imitation learning (IL) enables agents to mimic expert behavior without reward signals but faces challenges in cross-domain scenarios with high-dimensional, noisy, and incomplete visual observations. To address this, we propose Domain-Invariant Per-Frame Feature Extraction for Imitation Learning (DIFF-IL), a novel IL method that extracts domain-invariant features from individual frames and adapts them into sequences to isolate and replicate expert behaviors. We also introduce a frame-wise time labeling technique to segment expert behaviors by timesteps and assign rewards aligned with temporal contexts, enhancing task performance. Experiments across diverse visual environments demonstrate the effectiveness of DIFF-IL in addressing complex visual tasks., Comment: 8 pages main, 19 pages appendix with reference. Submitted to ICML 2025

Meta reinforcement learning aims to develop policies that generalize to unseen tasks sampled from a task distribution. While context-based meta-RL methods improve task representation using task latents, they often struggle with out-of-distribution (OOD) tasks. To address this, we propose Task-Aware Virtual Training (TAVT), a novel algorithm that accurately captures task characteristics for both training and OOD scenarios using metric-based representation learning. Our method successfully preserves task characteristics in virtual tasks and employs a state regularization technique to mitigate overestimation errors in state-varying environments. Numerical results demonstrate that TAVT significantly enhances generalization to OOD tasks across various MuJoCo and MetaWorld environments., Comment: 8 pages main paper, 19 pages appendices with reference, Submitted to ICML 2025

Designing Transformer architectures with the optimal layer normalization (LN) strategy that ensures large-scale training stability and expedite convergence has remained elusive, even in this era of large language models (LLMs). To this end, we present a comprehensive analytical foundation for understanding how different LN strategies influence training dynamics in large-scale Transformer training. Until recently, Pre-LN and Post-LN have long dominated standard practices despite their limitations in large-scale training. However, several open-source large-scale models have recently begun silently adopting a third strategy without much explanation. This strategy places layer normalization (LN) peripherally around sublayers, a design we term Peri-LN. While Peri-LN has demonstrated promising empirical performance, its precise mechanisms and benefits remain almost unexplored. Our in-depth analysis shows that Peri-LN strikes an ideal balance in variance growth -- unlike Pre-LN and Post-LN, which are prone to vanishing gradients and ``massive activations.'' To validate our theoretical insight, we conduct large-scale experiments on Transformers up to 3.2B parameters, showing that Peri-LN consistently achieves more balanced variance growth, steadier gradient flow, and convergence stability. Our results suggest that Peri-LN warrants broader consideration for large-scale Transformer architectures, providing renewed insights into the optimal placement and application of LN., Comment: Preprint

Vision sensors are extensively used for localizing a robot's pose, particularly in environments where global localization tools such as GPS or motion capture systems are unavailable. In many visual navigation systems, localization is achieved by detecting and tracking visual features or landmarks, which provide information about the sensor's relative pose. For reliable feature tracking and accurate pose estimation, it is crucial to maintain visibility of a sufficient number of features. This requirement can sometimes conflict with the robot's overall task objective. In this paper, we approach it as a constrained control problem. By leveraging the invariance properties of visibility constraints within the robot's kinematic model, we propose a real-time safety filter based on quadratic programming. This filter takes a reference velocity command as input and produces a modified velocity that minimally deviates from the reference while ensuring the information score from the currently visible features remains above a user-specified threshold. Numerical simulations demonstrate that the proposed safety filter preserves the invariance condition and ensures the visibility of more features than the required minimum. We also validated its real-world performance by integrating it into a visual simultaneous localization and mapping (SLAM) algorithm, where it maintained high estimation quality in challenging environments, outperforming a simple tracking controller., Comment: 7 pages, 6 figures, Accepted to 2025 IEEE International Conference on Robotics & Automation (ICRA 2025)

The introduction of 8-bit floating-point (FP8) computation units in modern AI accelerators has generated significant interest in FP8-based large language model (LLM) inference. Unlike 16-bit floating-point formats, FP8 in deep learning requires a shared scaling factor. Additionally, while E4M3 and E5M2 are well-defined at the individual value level, their scaling and accumulation methods remain unspecified and vary across hardware and software implementations. As a result, FP8 behaves more like a quantization format than a standard numeric representation. In this work, we provide the first comprehensive analysis of FP8 computation and acceleration on two AI accelerators: the NVIDIA H100 and Intel Gaudi 2. Our findings highlight that the Gaudi 2, by leveraging FP8, achieves higher throughput-to-power efficiency during LLM inference, offering valuable insights into the practical implications of FP8 adoption for datacenter-scale LLM serving.

While large language models (LLMs) excel at handling long-context sequences, they require substantial key-value (KV) caches to store contextual information, which can heavily burden computational efficiency and memory usage. Previous efforts to compress these KV caches primarily focused on reducing memory demands but were limited in enhancing latency. To address this issue, we introduce FastKV, a KV cache compression method designed to enhance latency for long-context sequences. To enhance processing speeds while maintaining accuracy, FastKV adopts a novel Token-Selective Propagation (TSP) approach that retains the full context information in the initial layers of LLMs and selectively propagates only a portion of this information in deeper layers even in the prefill stage. Additionally, FastKV incorporates grouped-query attention (GQA)-aware KV cache compression to exploit the advantages of GQA in both memory and computational efficiency. Our experimental results show that FastKV achieves 2.00$\times$ and 1.40$\times$ improvements in time-to-first-token (TTFT) and throughput, respectively, compared to HeadKV, the state-of-the-art KV cache compression method. Moreover, FastKV successfully maintains accuracy on long-context benchmarks at levels comparable to the baselines. Our code is available at https://github.com/dongwonjo/FastKV.

The upcoming all-sky infrared spectrophotometric SPHEREx mission is set to provide spatially resolved stellar mass maps of nearby galaxies, offering more detailed insights than integrated light observations. In this study, we develop a strategy for estimating stellar mass using SPHEREx by examining the dependence on different stellar population synthesis (SPS) models and proposing new scaling relations based on simulated SPHEREx data. We estimate the resolved stellar masses of 19 nearby late-type galaxies from the PHANGS-MUSE survey, treating these as fiducial masses. By testing four SPS models covering infrared wavelengths, i.e., E-MILES, Bruzual \& Charlot 2003 (BC03), Charlot \& Bruzual 2019 (CB19), and FSPS, we find systematic differences in mass-to-light ratios at $3.6~{\rm \mu m}$ ($M_{\ast}/L_{\rm 3.6\mu m}$) among the SPS models. In particular, BC03 and CB19 yield mass-to-light ratios on average $\sim0.2-0.3~{\rm dex}$ lower than those from E-MILES and FSPS. These mass-to-light ratios strongly correlate with stellar age, indicating a significant impact of young stellar populations on stellar mass measurements. Our analysis, incorporating fiducial masses and simulated SPHEREx data, identifies the $1.6~{\rm \mu m}$ band as the optimal wavelength for stellar mass estimation, with the lowest scatter ($0.15-0.20~{\rm dex}$) of the stellar mass. This scatter can be further reduced to $0.10-0.12~{\rm dex}$ across all SPS models by incorporating optical and SPHEREx colors. These results can provide guidance for measuring the stellar masses of the numerous nearby galaxies that SPHEREx will survey., Comment: Accepted for publication in AJ, 29 pages, 14 figures, 2 tables, 1 appendix

Recently, radars have been widely featured in robotics for their robustness in challenging weather conditions. Two commonly used radar types are spinning radars and phased-array radars, each offering distinct sensor characteristics. Existing datasets typically feature only a single type of radar, leading to the development of algorithms limited to that specific kind. In this work, we highlight that combining different radar types offers complementary advantages, which can be leveraged through a heterogeneous radar dataset. Moreover, this new dataset fosters research in multi-session and multi-robot scenarios where robots are equipped with different types of radars. In this context, we introduce the HeRCULES dataset, a comprehensive, multi-modal dataset with heterogeneous radars, FMCW LiDAR, IMU, GPS, and cameras. This is the first dataset to integrate 4D radar and spinning radar alongside FMCW LiDAR, offering unparalleled localization, mapping, and place recognition capabilities. The dataset covers diverse weather and lighting conditions and a range of urban traffic scenarios, enabling a comprehensive analysis across various environments. The sequence paths with multiple revisits and ground truth pose for each sensor enhance its suitability for place recognition research. We expect the HeRCULES dataset to facilitate odometry, mapping, place recognition, and sensor fusion research. The dataset and development tools are available at https://sites.google.com/view/herculesdataset., Comment: 2025 IEEE International Conference on Robotics and Automation (ICRA 2025)

$\chi$-separation is an advanced quantitative susceptibility mapping (QSM) method that is designed to generate paramagnetic ($\chi_{para}$) and diamagnetic ($|\chi_{dia}|$) susceptibility maps, reflecting the distribution of iron and myelin in the brain. However, vessels have shown artifacts, interfering with the accurate quantification of iron and myelin in applications. To address this challenge, a new vessel segmentation method for $\chi$-separation is developed. The method comprises three steps: 1) Seed generation from $\textit{R}_2^*$ and the product of $\chi_{para}$ and $|\chi_{dia}|$ maps; 2) Region growing, guided by vessel geometry, creating a vessel mask; 3) Refinement of the vessel mask by excluding non-vessel structures. The performance of the method was compared to conventional vessel segmentation methods both qualitatively and quantitatively. To demonstrate the utility of the method, it was tested in two applications: quantitative evaluation of a neural network-based $\chi$-separation reconstruction method ($\chi$-sepnet-$\textit{R}_2^*$) and population-averaged region of interest (ROI) analysis. The proposed method demonstrates superior performance to the conventional vessel segmentation methods, effectively excluding the non-vessel structures, achieving the highest Dice score coefficient. For the applications, applying vessel masks report notable improvements for the quantitative evaluation of $\chi$-sepnet-$\textit{R}_2^*$ and statistically significant differences in population-averaged ROI analysis. These applications suggest excluding vessels when analyzing the $\chi$-separation maps provide more accurate evaluations. The proposed method has the potential to facilitate various applications, offering reliable analysis through the generation of a high-quality vessel mask.

We perform an MMT/Hectospec redshift survey of the North Ecliptic Pole Wide (NEPW) field covering 5.4 square degrees, and use it to estimate the photometric redshifts for the sources without spectroscopic redshifts. By combining 2572 newly measured redshifts from our survey with existing data from the literature, we create a large sample of 4421 galaxies with spectroscopic redshifts in the NEPW field. Using this sample, we estimate photometric redshifts of 77755 sources in the band-merged catalog of the NEPW field with a random forest model. The estimated photometric redshifts are generally consistent with the spectroscopic redshifts, with a dispersion of 0.028, an outlier fraction of 7.3%, and a bias of -0.01. We find that the standard deviation of the prediction from each decision tree in the random forest model can be used to infer the fraction of catastrophic outliers and the measurement uncertainties. We test various combinations of input observables, including colors and magnitude uncertainties, and find that the details of these various combinations do not change the prediction accuracy much. As a result, we provide a catalog of 77755 sources in the NEPW field, which includes both spectroscopic and photometric redshifts up to z~2. This dataset has significant legacy value for studies in the NEPW region, especially with upcoming space missions such as JWST, Euclid, and SPHEREx., Comment: 18 pagees, 12 figures, 4 tables

Ensuring fairness in medical image segmentation is critical due to biases in imbalanced clinical data acquisition caused by demographic attributes (e.g., age, sex, race) and clinical factors (e.g., disease severity). To address these challenges, we introduce Distribution-aware Mixture of Experts (dMoE), inspired by optimal control theory. We provide a comprehensive analysis of its underlying mechanisms and clarify dMoE's role in adapting to heterogeneous distributions in medical image segmentation. Furthermore, we integrate dMoE into multiple network architectures, demonstrating its broad applicability across diverse medical image analysis tasks. By incorporating demographic and clinical factors, dMoE achieves state-of-the-art performance on two 2D benchmark datasets and a 3D in-house dataset. Our results highlight the effectiveness of dMoE in mitigating biases from imbalanced distributions, offering a promising approach to bridging control theory and medical image segmentation within fairness learning paradigms. The source code will be made available., Comment: 12 pages, 3 figures, 9 tables

Sentiment analysis of alternative tobacco products on social media is important for tobacco control research. Large Language Models (LLMs) can help streamline the labor-intensive human sentiment analysis process. This study examined the accuracy of LLMs in replicating human sentiment evaluation of social media messages about heated tobacco products (HTPs). The research used GPT-3.5 and GPT-4 Turbo to classify 500 Facebook and 500 Twitter messages, including anti-HTPs, pro-HTPs, and neutral messages. The models evaluated each message up to 20 times, and their majority label was compared to human evaluators. Results showed that GPT-3.5 accurately replicated human sentiment 61.2% of the time for Facebook messages and 57.0% for Twitter messages. GPT-4 Turbo performed better, with 81.7% accuracy for Facebook and 77.0% for Twitter. Using three response instances, GPT-4 Turbo achieved 99% of the accuracy of twenty instances. GPT-4 Turbo also had higher accuracy for anti- and pro-HTPs messages compared to neutral ones. Misclassifications by GPT-3.5 often involved anti- or pro-HTPs messages being labeled as neutral or irrelevant, while GPT-4 Turbo showed improvements across all categories. In conclusion, LLMs can be used for sentiment analysis of HTP-related social media messages, with GPT-4 Turbo reaching around 80% accuracy compared to human experts. However, there's a risk of misrepresenting overall sentiment due to differences in accuracy across sentiment categories.

Under-Display Camera (UDC) houses a digital camera lens under a display panel. However, UDC introduces complex degradations such as noise, blur, decrease in transmittance, and flare. Despite the remarkable progress, previous research on UDC mainly focuses on eliminating diffraction in the spatial domain and rarely explores its potential in the frequency domain. It is essential to consider both the spatial and frequency domains effectively. For example, degradations, such as noise and blur, can be addressed by local information (e.g., CNN kernels in the spatial domain). At the same time, tackling flares may require leveraging global information (e.g., the frequency domain). In this paper, we revisit the UDC degradations in the Fourier space and figure out intrinsic frequency priors that imply the presence of the flares. Based on this observation, we propose a novel multi-level DNN architecture called SFIM. It efficiently restores UDC-distorted images by integrating local and global (the collective contribution of all points in the image) information. The architecture exploits CNNs to capture local information and FFT-based models to capture global information. SFIM comprises a spatial domain block (SDB), a Frequency Domain Block (FDB), and an Attention-based Multi-level Integration Block (AMIB). Specifically, SDB focuses more on detailed textures such as noise and blur, FDB emphasizes irregular texture loss in extensive areas such as flare, and AMIB enables effective cross-domain interaction. SFIM's superior performance over state-of-the-art approaches is demonstrated through rigorous quantitative and qualitative assessments across three UDC benchmarks., Comment: Main body (10 pages, 9 Figures, 5 Tables), References (3 pages), Appendix (8 pages, 6 Figures, 6 Tables)

In this paper, we develop a neural network-based approach for time-series prediction in unknown Hamiltonian dynamical systems. Our approach leverages a surrogate model and learns the system dynamics using generalized coordinates (positions) and their conjugate momenta while preserving a constant Hamiltonian. To further enhance long-term prediction accuracy, we introduce an Autoregressive Hamiltonian Neural Network, which incorporates autoregressive prediction errors into the training objective. Additionally, we employ Bayesian data assimilation to refine predictions in real-time using online measurement data. Numerical experiments on a spring-mass system and highly elliptic orbits under gravitational perturbations demonstrate the effectiveness of the proposed method, highlighting its potential for accurate and robust long-term predictions., Comment: 8 pages, 12 figures

As one of the most promising methods in self-supervised learning, contrastive learning has achieved a series of breakthroughs across numerous fields. A predominant approach to implementing contrastive learning is applying InfoNCE loss: By capturing the similarities between pairs, InfoNCE loss enables learning the representation of data. Albeit its success, adopting InfoNCE loss requires tuning a temperature, which is a core hyperparameter for calibrating similarity scores. Despite its significance and sensitivity to performance being emphasized by several studies, searching for a valid temperature requires extensive trial-and-error-based experiments, which increases the difficulty of adopting InfoNCE loss. To address this difficulty, we propose a novel method to deploy InfoNCE loss without temperature. Specifically, we replace temperature scaling with the inverse hyperbolic tangent function, resulting in a modified InfoNCE loss. In addition to hyperparameter-free deployment, we observed that the proposed method even yielded a performance gain in contrastive learning. Our detailed theoretical analysis discovers that the current practice of temperature scaling in InfoNCE loss causes serious problems in gradient descent, whereas our method provides desirable gradient properties. The proposed method was validated on five benchmarks on contrastive learning, yielding satisfactory results without temperature tuning., Comment: 10 pages, 5 figures

Graph-structured data plays a vital role in numerous domains, such as social networks, citation networks, commonsense reasoning graphs and knowledge graphs. While graph neural networks have been employed for graph processing, recent advancements have explored integrating large language models for graph-based tasks. In this paper, we propose a novel approach named Learnable Graph Pooling Token (LGPT), which addresses the limitations of the scalability issues in node-level projection and information loss in graph-level projection. LGPT enables flexible and efficient graph representation by introducing learnable parameters that act as tokens in large language models, balancing fine-grained and global graph information. Additionally, we investigate an Early Query Fusion technique, which fuses query context before constructing the graph representation, leading to more effective graph embeddings. Our method achieves a 4.13\% performance improvement on the GraphQA benchmark without training the large language model, demonstrating significant gains in handling complex textual-attributed graph data.

The gravitational wave (GW) event S230518h is a potential binary neutron star-black hole merger (NSBH) event that was detected during engineering run 15 (ER15), which served as the commissioning period before the LIGO-Virgo-KAGRA (LVK) O4a observing run. Despite its low probability of producing detectable electromagnetic emissions, we performed extensive follow-up observations of this event using the GECKO telescopes in the southern hemisphere. Our observation covered 61.7\% of the 90\% credible region, a $\rm 284\:deg^2$ area accessible from the southern hemisphere, reaching a median limiting magnitude of $R=21.6$ mag. In these images, we conducted a systematic search for an optical counterpart of this event by combining a CNN-based classifier and human verification. We identified 128 transient candidates, but no significant optical counterpart was found that could have caused the GW signal. Furthermore, we provide feasible KN properties that are consistent with the upper limits of observation. Although no optical counterpart was found, our result demonstrates both GECKO's efficient wide-field follow-up capabilities and usefulness for constraining properties of kilonovae from NSBH mergers at distances of $\sim 200$ Mpc., Comment: 25 pages, 14 figures, Accepted for publication in ApJ