Search

Recently, crossbar array based in-memory accelerators have been gaining interest due to their high throughput and energy efficiency. While software and compiler support for the in-memory accelerators has also been introduced, they are currently limited to the case where all weights are assumed to be on-chip. This limitation becomes apparent with the significantly increasing network sizes compared to the in-memory footprint. Weight replacement schemes are essential to address this issue. We propose COMPASS, a compiler framework for resource-constrained crossbar-based processing-in-memory (PIM) deep neural network (DNN) accelerators. COMPASS is specially targeted for networks that exceed the capacity of PIM crossbar arrays, necessitating access to external memories. We propose an algorithm to determine the optimal partitioning that divides the layers so that each partition can be accelerated on chip. Our scheme takes into account the data dependence between layers, core utilization, and the number of write instructions to minimize latency, memory accesses, and improve energy efficiency. Simulation results demonstrate that COMPASS can accommodate much more networks using a minimal memory footprint, while improving throughput by 1.78X and providing 1.28X savings in energy-delay product (EDP) over baseline partitioning methods., Comment: Accepted IEEE DATE 2025

Chirality, an intrinsic preference for a specific handedness, is a fundamental characteristic observed in nature. In magnetism, magnetic chirality arises from the anti-symmetric Dzyaloshinskii-Moriya interaction in competition with the symmetric Heisenberg exchange interaction. Traditionally, the anti-symmetric interaction has been considered minor relative to the symmetric interaction. In this study, we demonstrate an observation of giant magnetic chirality during the dimensionality crossover of magnetic materials from three-dimensional to two-dimensional. The ratio between the anti-symmetric and symmetric interactions exhibits a reversal in their dominance over this crossover, overturning the traditional consideration. This observation is validated theoretically using a non-local interaction model and tight-binding calculation with distinct pairing schemes for each exchange interaction throughout the crossover. Additional experiments investigating the asphericity of orbital moments corroborate the robustness of our findings. Our findings highlight the critical role of dimensionality in shaping magnetic chirality and offer strategies for engineering chiral magnet states with unprecedented strength, desired for the design of spintronic materials., Comment: 21 pages, 4 figures

We show that a minimal surface meeting a sphere at a 90-degree angle can be reflected across the sphere. Using this reflection, we prove the uniqueness that every embedded free boundary minimal annulus in a ball is necessarily the critical catenoid., Comment: 16 pages

Existing open-vocabulary object detection (OVD) develops methods for testing unseen categories by aligning object region embeddings with corresponding VLM features. A recent study leverages the idea that VLMs implicitly learn compositional structures of semantic concepts within the image. Instead of using an individual region embedding, it utilizes a bag of region embeddings as a new representation to incorporate compositional structures into the OVD task. However, this approach often fails to capture the contextual concepts of each region, leading to noisy compositional structures. This results in only marginal performance improvements and reduced efficiency. To address this, we propose a novel concept-based alignment method that samples a more powerful and efficient compositional structure. Our approach groups contextually related ``concepts'' into a bag and adjusts the scale of concepts within the bag for more effective embedding alignment. Combined with Faster R-CNN, our method achieves improvements of 2.6 box AP50 and 0.5 mask AP over prior work on novel categories in the open-vocabulary COCO and LVIS benchmarks. Furthermore, our method reduces CLIP computation in FLOPs by 80.3% compared to previous research, significantly enhancing efficiency. Experimental results demonstrate that the proposed method outperforms previous state-of-the-art models on the OVD datasets., Comment: 19 pages

To deploy LLMs on resource-contained platforms such as mobile robotics and wearables, non-transformers LLMs have achieved major breakthroughs. Recently, a novel RNN-based LLM family, Repentance Weighted Key Value (RWKV) models have shown promising results in text generation on resource-constrained devices thanks to their computational efficiency. However, these models remain too large to be deployed on embedded devices due to their high parameter count. In this paper, we propose an efficient suite of compression techniques, tailored to the RWKV architecture. These techniques include low-rank approximation, sparsity predictors, and clustering head, designed to align with the model size. Our methods compress the RWKV models by 4.95--3.8x with only 2.95pp loss in accuracy.

We propose an efficient radiance field rendering algorithm that incorporates a rasterization process on sparse voxels without neural networks or 3D Gaussians. There are two key contributions coupled with the proposed system. The first is to render sparse voxels in the correct depth order along pixel rays by using dynamic Morton ordering. This avoids the well-known popping artifact found in Gaussian splatting. Second, we adaptively fit sparse voxels to different levels of detail within scenes, faithfully reproducing scene details while achieving high rendering frame rates. Our method improves the previous neural-free voxel grid representation by over 4db PSNR and more than 10x rendering FPS speedup, achieving state-of-the-art comparable novel-view synthesis results. Additionally, our neural-free sparse voxels are seamlessly compatible with grid-based 3D processing algorithms. We achieve promising mesh reconstruction accuracy by integrating TSDF-Fusion and Marching Cubes into our sparse grid system., Comment: Code release in progress

Open-vocabulary segmentation (OVS) has gained attention for its ability to recognize a broader range of classes. However, OVS models show significant performance drops when applied to unseen domains beyond the previous training dataset. Fine-tuning these models on new datasets can improve performance, but often leads to the catastrophic forgetting of previously learned knowledge. To address this issue, we propose a method that allows OVS models to learn information from new domains while preserving prior knowledge. Our approach begins by evaluating the input sample's proximity to multiple domains, using precomputed multivariate normal distributions for each domain. Based on this prediction, we dynamically interpolate between the weights of the pre-trained decoder and the fine-tuned decoders. Extensive experiments demonstrate that this approach allows OVS models to adapt to new domains while maintaining performance on the previous training dataset. The source code is available at https://github.com/dongjunhwang/dwi.

Single spot radiometers are widely used to estimate temperature of body heated by high temperature, specially blast furnace, glass fusion, boiler of power plant, and so on. In previous papers on single spot radiometer a heated body was deduced according to interpolation functions between heated body temperature and photocurrent, which had previously measured. However, fitting functions required for current-to-voltage converter, non-inverting amplifier, and logarithmic amplifier have never reported. Therefore, in this paper the fitting functions are reported that can obtain the relationship between radiance temperatures of a body and outputs of amplifiers in single spot radiometer., Comment: 11 pages, 9 figures

Binary Code Similarity Analysis (BCSA) has a wide spectrum of applications, including plagiarism detection, vulnerability discovery, and malware analysis, thus drawing significant attention from the security community. However, conventional techniques often face challenges in balancing both accuracy and scalability simultaneously. To overcome these existing problems, a surge of deep learning-based work has been recently proposed. Unfortunately, many researchers still find it extremely difficult to conduct relevant studies or extend existing approaches. First, prior work typically relies on proprietary benchmark without making the entire dataset publicly accessible. Consequently, a large-scale, well-labeled dataset for binary code similarity analysis remains precious and scarce. Moreover, previous work has primarily focused on comparing at the function level, rather than exploring other finer granularities. Therefore, we argue that the lack of a fine-grained dataset for BCSA leaves a critical gap in current research. To address these challenges, we construct a benchmark dataset for fine-grained binary code similarity analysis called BinSimDB, which contains equivalent pairs of smaller binary code snippets, such as basic blocks. Specifically, we propose BMerge and BPair algorithms to bridge the discrepancies between two binary code snippets caused by different optimization levels or platforms. Furthermore, we empirically study the properties of our dataset and evaluate its effectiveness for the BCSA research. The experimental results demonstrate that BinSimDB significantly improves the performance of binary code similarity comparison., Comment: Published by International Conference on Security and Privacy in Communication Networks (SecureComm 2024)

Machine unlearning aims to selectively remove specific knowledge from a model. Current methods, such as task arithmetic, rely on fine-tuning models on the forget set, generating a task vector, and subtracting it from the original model. However, we argue the effectiveness of this approach is highly sensitive to hyperparameter selection, necessitating careful validation to identify the best model among many fine-tuned candidates. In this paper, we propose a novel method that leverages all given fine-tuned models rather than selecting a single one. By constructing task vectors from models trained with varied hyperparameters and merging only the components of the task vectors with consistent signs, we perform unlearning by negating the merged task vector from the original model. Given that existing methods also utilize multiple fine-tuned models, our approach delivers more effective unlearning without incurring additional computational costs. We demonstrate the effectiveness of our method on both vision-language models and standard image classification models, showing improved unlearning performance with minimal degradation on the retain set, outperforming state-of-the-art techniques.

We investigate the Lyman-$\alpha$ (Ly$\alpha$) and Lyman continuum (LyC) properties of the Sunburst Arc, a $z=2.37$ gravitationally lensed galaxy with a multiply-imaged, compact region leaking LyC and a triple-peaked Ly$\alpha$ profile indicating direct Ly$\alpha$ escape. Non-LyC-leaking regions show a redshifted Ly$\alpha$ peak, a redshifted and central Ly$\alpha$ peak, or a triple-peaked Ly$\alpha$ profile. We measure the properties of the Ly$\alpha$ profile from different regions of the galaxy using $R\sim5000$ Magellan/MagE spectra. We compare the Ly$\alpha$ spectral properties to LyC and narrowband Ly$\alpha$ maps from Hubble Space Telescope (HST) imaging to explore the subgalactic Ly$\alpha-$LyC connection. We find strong correlations (Pearson correlation coefficient $r>0.6$) between the LyC escape fraction ($f_{\rm esc}^{\rm LyC}$) and Ly$\alpha$ (1) peak separation $v_{\rm{sep}}$, (2) ratio of the minimum flux density between the redshifted and blueshifted Ly$\alpha$ peaks to continuum flux density $f_{\rm{min}}/f_{\rm{cont}}$, and (3) equivalent width. We favor a complex \ion{H}{1} geometry to explain the Ly$\alpha$ profiles from non-LyC-leaking regions and suggest two \ion{H}{1} geometries that could diffuse and/or rescatter the central Ly$\alpha$ peak from the LyC-leaking region into our sightline across transverse distances of several hundred parsecs. Our results emphasize the complexity of Ly$\alpha$ radiative transfer and its sensitivity to the anisotropies of \ion{H}{1} gas on subgalactic scales. Large differences in the physical scales on which we observe spatially variable direct escape Ly$\alpha$, blueshifted Ly$\alpha$, and escaping LyC photons in the Sunburst Arc underscore the importance of resolving the physical scales that govern Ly$\alpha$ and LyC escape., Comment: Submitted to The Astrophysical Journal with revisions from the first referee report. Comments welcome

Federated Learning is a promising approach for training machine learning models while preserving data privacy, but its distributed nature makes it vulnerable to backdoor attacks, particularly in NLP tasks while related research remains limited. This paper introduces SDBA, a novel backdoor attack mechanism designed for NLP tasks in FL environments. Our systematic analysis across LSTM and GPT-2 models identifies the most vulnerable layers for backdoor injection and achieves both stealth and long-lasting durability through layer-wise gradient masking and top-k% gradient masking within these layers. Experiments on next token prediction and sentiment analysis tasks show that SDBA outperforms existing backdoors in durability and effectively bypasses representative defense mechanisms, with notable performance in LLM such as GPT-2. These results underscore the need for robust defense strategies in NLP-based FL systems., Comment: 13 pages, 13 figures This work has been submitted to the IEEE for possible publication

The SoccerNet 2024 challenges represent the fourth annual video understanding challenges organized by the SoccerNet team. These challenges aim to advance research across multiple themes in football, including broadcast video understanding, field understanding, and player understanding. This year, the challenges encompass four vision-based tasks. (1) Ball Action Spotting, focusing on precisely localizing when and which soccer actions related to the ball occur, (2) Dense Video Captioning, focusing on describing the broadcast with natural language and anchored timestamps, (3) Multi-View Foul Recognition, a novel task focusing on analyzing multiple viewpoints of a potential foul incident to classify whether a foul occurred and assess its severity, (4) Game State Reconstruction, another novel task focusing on reconstructing the game state from broadcast videos onto a 2D top-view map of the field. Detailed information about the tasks, challenges, and leaderboards can be found at https://www.soccer-net.org, with baselines and development kits available at https://github.com/SoccerNet., Comment: 7 pages, 1 figure

This report presents a solution for the swing-up and stabilisation tasks of the acrobot and the pendubot, developed for the AI Olympics competition at IROS 2024. Our approach employs the Average-Reward Entropy Advantage Policy Optimization (AR-EAPO), a model-free reinforcement learning (RL) algorithm that combines average-reward RL and maximum entropy RL. Results demonstrate that our controller achieves improved performance and robustness scores compared to established baseline methods in both the acrobot and pendubot scenarios, without the need for a heavily engineered reward function or system model. The current results are applicable exclusively to the simulation stage setup.

Recent Vision Transformer (ViT)-based methods for Image Super-Resolution have demonstrated impressive performance. However, they suffer from significant complexity, resulting in high inference times and memory usage. Additionally, ViT models using Window Self-Attention (WSA) face challenges in processing regions outside their windows. To address these issues, we propose the Low-to-high Multi-Level Transformer (LMLT), which employs attention with varying feature sizes for each head. LMLT divides image features along the channel dimension, gradually reduces spatial size for lower heads, and applies self-attention to each head. This approach effectively captures both local and global information. By integrating the results from lower heads into higher heads, LMLT overcomes the window boundary issues in self-attention. Extensive experiments show that our model significantly reduces inference time and GPU memory usage while maintaining or even surpassing the performance of state-of-the-art ViT-based Image Super-Resolution methods. Our codes are availiable at https://github.com/jwgdmkj/LMLT.

Hallucinations in Multimodal Large Language Models (MLLMs) where generated responses fail to accurately reflect the given image pose a significant challenge to their reliability. To address this, we introduce ConVis, a novel training-free contrastive decoding method. ConVis leverages a text-to-image (T2I) generation model to semantically reconstruct the given image from hallucinated captions. By comparing the contrasting probability distributions produced by the original and reconstructed images, ConVis enables MLLMs to capture visual contrastive signals that penalize hallucination generation. Notably, this method operates purely within the decoding process, eliminating the need for additional data or model updates. Our extensive experiments on five popular benchmarks demonstrate that ConVis effectively reduces hallucinations across various MLLMs, highlighting its potential to enhance model reliability., Comment: First two authors contributed equally. Source code is available at https://github.com/yejipark-m/ConVis

Artificial intelligence applied to retinal images offers significant potential for recognizing signs and symptoms of retinal conditions and expediting the diagnosis of eye diseases and systemic disorders. However, developing generalized artificial intelligence models for medical data often requires a large number of labeled images representing various disease signs, and most models are typically task-specific, focusing on major retinal diseases. In this study, we developed a Fundus-Specific Pretrained Model (Image+Fundus), a supervised artificial intelligence model trained to detect abnormalities in fundus images. A total of 57,803 images were used to develop this pretrained model, which achieved superior performance across various downstream tasks, indicating that our proposed model outperforms other general methods. Our Image+Fundus model offers a generalized approach to improve model performance while reducing the number of labeled datasets required. Additionally, it provides more disease-specific insights into fundus images, with visualizations generated by our model. These disease-specific foundation models are invaluable in enhancing the performance and efficiency of deep learning models in the field of fundus imaging., Comment: 10 pages, 4 figures

Entropy Regularisation is a widely adopted technique that enhances policy optimisation performance and stability. A notable form of entropy regularisation is augmenting the objective with an entropy term, thereby simultaneously optimising the expected return and the entropy. This framework, known as maximum entropy reinforcement learning (MaxEnt RL), has shown theoretical and empirical successes. However, its practical application in straightforward on-policy actor-critic settings remains surprisingly underexplored. We hypothesise that this is due to the difficulty of managing the entropy reward in practice. This paper proposes a simple method of separating the entropy objective from the MaxEnt RL objective, which facilitates the implementation of MaxEnt RL in on-policy settings. Our empirical evaluations demonstrate that extending Proximal Policy Optimisation (PPO) and Trust Region Policy Optimisation (TRPO) within the MaxEnt framework improves policy optimisation performance in both MuJoCo and Procgen tasks. Additionally, our results highlight MaxEnt RL's capacity to enhance generalisation.

The AMoRE collaboration searches for neutrinoless double beta decay of $^{100}$Mo using molybdate scintillating crystals via low temperature thermal calorimetric detection. The early phases of the experiment, AMoRE-pilot and AMoRE-I, have demonstrated competitive discovery potential. Presently, the AMoRE-II experiment, featuring a large detector array with about 90 kg of $^{100}$Mo isotope, is under construction.This paper discusses the baseline design and characterization of the lithium molybdate cryogenic calorimeters to be used in the AMoRE-II detector modules. The results from prototype setups that incorporate new housing structures and two different crystal masses (316 g and 517 - 521 g), operated at 10 mK temperature, show energy resolutions (FWHM) of 7.55 - 8.82 keV at the 2.615 MeV $^{208}$Tl $\gamma$ line, and effective light detection of 0.79 - 0.96 keV/MeV. The simultaneous heat and light detection enables clear separation of alpha particles with a discrimination power of 12.37 - 19.50 at the energy region around $^6$Li(n, $\alpha$)$^3$H with Q-value = 4.785 MeV. Promising detector performances were demonstrated at temperatures as high as 30 mK, which relaxes the temperature constraints for operating the large AMoRE-II array.

AMoRE searches for the signature of neutrinoless double beta decay of $^{100}$Mo with a 100 kg sample of enriched $^{100}$Mo. Scintillating molybdate crystals coupled with a metallic magnetic calorimeter operate at milli-Kelvin temperatures to measure the energy of electrons emitted in the decay. As a demonstration of the full-scale AMoRE, we conducted AMoRE-I, a pre-experiment with 18 molybdate crystals, at the Yangyang Underground Laboratory for over two years. The exposure was 8.02 kg$\cdot$year (or 3.89 kg$_{\mathrm{^{100}Mo}}\cdot$year) and the total background rate near the Q-value was 0.025 $\pm$ 0.002 counts/keV/kg/year. We observed no indication of $0\nu\beta\beta$ decay and report a new lower limit of the half-life of $^{100}$Mo $0\nu\beta\beta$ decay as $ T^{0\nu}_{1/2}>3.0\times10^{24}~\mathrm{years}$ at 90\% confidence level. The effective Majorana mass limit range is $m_{\beta\beta}<$(210--610) meV using nuclear matrix elements estimated in the framework of different models, including the recent shell model calculations., Comment: 8 pages, 5 figures

UWB has recently gained new attention as an auxiliary sensor in the field of robot localization due to its compactness and ease of distance measurement. Consequently, various UWB-related localization and dataset research have increased. Despite this broad interest, there is a lack of UWB datasets that thoroughly analyze the performance of UWB ranging measurement. To address this issue, our paper introduces a UWB dataset that examines UWB relative pose factors affecting ranging measurement. To the best of our knowledge, our dataset is the first to analyze these factors while rigorously providing precise ground-truth UWB poses. The dataset is accessible at https://github.com/cjhhalla/RCV_uwb_dataset ., Comment: 4 pages

We propose a metasurface antenna capable of real time holographic beam steering. An array of reconfigurable dipoeles can generate on demand far field patterns of radiation through the specific encoding of meta atomic states. i.e., the configuration of each dipole. Suitable states for the generation of the desired patterns can be identified using iteartion, but this is very slow and needs to be done for each far field pattern. Here, we present a deep learning based method for the control of a metasurface antenna with point dipole elements that vary in their state using dipole polarizability. Instead of iteration, we adopt a deep learning algorithm that combines an autoencoder with an electromagnetic scattering equation to determin the states required for a target far field pattern in real time. The scattering equation from Born approximation is used as the decoder in training the neural network, and analytic Green's function calculation is used to check the validity of Born approximation. Our learning based algorithm requires a computing time of within in 200 microseconds to determine the meta atomic states, thus enabling the real time opeartion of a holographic antenna.

AMoRE-II aims to search for neutrinoless double beta decay with an array of 423 Li$_2$$^{100}$MoO$_4$ crystals operating in the cryogenic system as the main phase of the Advanced Molybdenum-based Rare process Experiment (AMoRE). AMoRE has been planned to operate in three phases: AMoRE-pilot, AMoRE-I, and AMoRE-II. AMoRE-II is currently being installed at the Yemi Underground Laboratory, located approximately 1000 meters deep in Jeongseon, Korea. The goal of AMoRE-II is to reach up to $T^{0\nu\beta\beta}_{1/2}$ $\sim$ 6 $\times$ 10$^{26}$ years, corresponding to an effective Majorana mass of 15 - 29 meV, covering all the inverted mass hierarchy regions. To achieve this, the background level of the experimental configurations and possible background sources of gamma and beta events should be well understood. We have intensively performed Monte Carlo simulations using the GEANT4 toolkit in all the experimental configurations with potential sources. We report the estimated background level that meets the 10$^{-4}$counts/(keV$\cdot$kg$\cdot$yr) requirement for AMoRE-II in the region of interest (ROI) and show the projected half-life sensitivity based on the simulation study.

We present Meta MMO, a collection of many-agent minigames for use as a reinforcement learning benchmark. Meta MMO is built on top of Neural MMO, a massively multiagent environment that has been the subject of two previous NeurIPS competitions. Our work expands Neural MMO with several computationally efficient minigames. We explore generalization across Meta MMO by learning to play several minigames with a single set of weights. We release the environment, baselines, and training code under the MIT license. We hope that Meta MMO will spur additional progress on Neural MMO and, more generally, will serve as a useful benchmark for many-agent generalization.

The linear representation hypothesis is the informal idea that semantic concepts are encoded as linear directions in the representation spaces of large language models (LLMs). Previous work has shown how to make this notion precise for representing binary concepts that have natural contrasts (e.g., {male, female}) as directions in representation space. However, many natural concepts do not have natural contrasts (e.g., whether the output is about an animal). In this work, we show how to extend the formalization of the linear representation hypothesis to represent features (e.g., is_animal) as vectors. This allows us to immediately formalize the representation of categorical concepts as polytopes in the representation space. Further, we use the formalization to prove a relationship between the hierarchical structure of concepts and the geometry of their representations. We validate these theoretical results on the Gemma and LLaMA-3 large language models, estimating representations for 900+ hierarchically related concepts using data from WordNet., Comment: Best Paper Award at the ICML 2024 Workshop on Mechanistic Interpretability. Code is available at https://github.com/KihoPark/LLM_Categorical_Hierarchical_Representations

Unsupervised domain adaptation (UDA) for semantic segmentation aims to transfer the pixel-wise knowledge from the labeled source domain to the unlabeled target domain. However, current UDA methods typically assume a shared label space between source and target, limiting their applicability in real-world scenarios where novel categories may emerge in the target domain. In this paper, we introduce Open-Set Domain Adaptation for Semantic Segmentation (OSDA-SS) for the first time, where the target domain includes unknown classes. We identify two major problems in the OSDA-SS scenario as follows: 1) the existing UDA methods struggle to predict the exact boundary of the unknown classes, and 2) they fail to accurately predict the shape of the unknown classes. To address these issues, we propose Boundary and Unknown Shape-Aware open-set domain adaptation, coined BUS. Our BUS can accurately discern the boundaries between known and unknown classes in a contrastive manner using a novel dilation-erosion-based contrastive loss. In addition, we propose OpenReMix, a new domain mixing augmentation method that guides our model to effectively learn domain and size-invariant features for improving the shape detection of the known and unknown classes. Through extensive experiments, we demonstrate that our proposed BUS effectively detects unknown classes in the challenging OSDA-SS scenario compared to the previous methods by a large margin. The code is available at https://github.com/KHU-AGI/BUS., Comment: 14 pages, 5 figures, 13 tables, CVPR 2024 Poster

Large language models (LLMs) are trained on a vast amount of human-written data, but data providers often remain uncredited. In response to this issue, data valuation (or data attribution), which quantifies the contribution or value of each data to the model output, has been discussed as a potential solution. Nevertheless, applying existing data valuation methods to recent LLMs and their vast training datasets has been largely limited by prohibitive compute and memory costs. In this work, we focus on influence functions, a popular gradient-based data valuation method, and significantly improve its scalability with an efficient gradient projection strategy called LoGra that leverages the gradient structure in backpropagation. We then provide a theoretical motivation of gradient projection approaches to influence functions to promote trust in the data valuation process. Lastly, we lower the barrier to implementing data valuation systems by introducing LogIX, a software package that can transform existing training code into data valuation code with minimal effort. In our data valuation experiments, LoGra achieves competitive accuracy against more expensive baselines while showing up to 6,500x improvement in throughput and 5x reduction in GPU memory usage when applied to Llama3-8B-Instruct and the 1B-token dataset.

Visualizing citation relations with network structures is widely used, but the visual complexity can make it challenging for individual researchers trying to navigate them. We collected data from 18 researchers with an interface that we designed using network simplification methods and analyzed how users browsed and identified important papers. Our analysis reveals six major patterns used for identifying papers of interest, which can be categorized into three key components: Fields, Bridges, and Foundations, each viewed from two distinct perspectives: layout-oriented and connection-oriented. The connection-oriented approach was found to be more reliable for selecting relevant papers, but the layout-oriented method was adopted more often, even though it led to unexpected results and user frustration. Our findings emphasize the importance of integrating these components and the necessity to balance visual layouts with meaningful connections to enhance the effectiveness of citation networks in academic browsing systems.

"Godzilla" is a peculiar object within the gravitationally lensed Sunburst Arc at $z=2.37$. Despite being very bright, it appears in only one of the twelve lensed images of the source galaxy, and shows exotic spectroscopic properties not found in any other clumps. We use JWST's unique combination of spatial resolution and spectroscopic sensitivity to provide a unified, coherent explanation of the physical nature of Godzilla. We measure fluxes and kinematic properties of rest-optical emission lines in Godzilla and surrounding regions. Using standard line ratio-based diagnostic methods in combination with NIRCam imaging and ground based rest-UV spectra, we characterize Godzilla and its surroundings. Among around 60 detected lines, we find a cascade of strong O I lines pumped by intense Ly$\beta$ emission, as well as Ly$\alpha$-pumped rest-optical Fe II lines, reminiscent of the Weigelt blobs in the local LBV star Eta Carinae. Spectra and images of Godzilla and two faint adjacent images, and the detection of a low-surface brightness foreground galaxy in the NIRCam data, support the interpretation that Godzilla is an extremely magnified object due to the alignment with lensing caustics. We find that Godzilla is part of a previously identified clump, comprising $\sim10 - 25$ % of it, with magnifications in the range of $\approx600-25,000$ depending on the models and images in comparison. The unique O I source in Godzilla is well explained by a non-erupting LBV accompanied by a hotter companion and/or gas condensations exposed to more intense radiation compared to the Weigelt blobs. If Godzilla is confirmed to contain an LBV star, it expands the distance to the furthest known LBV from a dozen Mpc to several Gpc., Comment: 22 pages, 19 figures. Submitted to A&A