Search

The Wallace Foundation commissioned Research for Action and McClanahan Associates to study how out-of-school time intermediaries (OSTIs) responded to the COVID-19 pandemic and the role they could play in supporting communities moving forward. Citywide out-of-school time (OST) systems encompass a wide variety of afterschool and summer programs and associated provider organizations, youth participants, and their families. These systems were stretched during the COVID-19 pandemic and the national racial reckoning to respond to community needs in new and more equitable ways. Also impacted were OSTIs that manage OST systems within individual cities. OSTIs operate through various governance structures and perform a range of coordination functions intended to ensure quality, access, and equity in OST systems so that all youth have access to the unique out-of-school resources and opportunities they need to thrive.

In-context learning (ICL) is a powerful paradigm where large language models (LLMs) benefit from task demonstrations added to the prompt. Yet, selecting optimal demonstrations is not trivial, especially for complex or multi-modal tasks where input and output distributions differ. We hypothesize that forming task-specific representations of the input is key. In this paper, we propose a method to align representations of natural language questions and those of SQL queries in a shared embedding space. Our technique, dubbed MARLO - Metadata-Agnostic Representation Learning for Text-tO-SQL - uses query structure to model querying intent without over-indexing on underlying database metadata (i.e. tables, columns, or domain-specific entities of a database referenced in the question or query). This allows MARLO to select examples that are structurally and semantically relevant for the task rather than examples that are spuriously related to a certain domain or question phrasing. When used to retrieve examples based on question similarity, MARLO shows superior performance compared to generic embedding models (on average +2.9\%pt. in execution accuracy) on the Spider benchmark. It also outperforms the next best method that masks metadata information by +0.8\%pt. in execution accuracy on average, while imposing a significantly lower inference latency., Comment: Accepted to NeurIPS 2024 Table Representation Learning workshop

KnowWhereGraph is one of the largest fully publicly available geospatial knowledge graphs. It includes data from 30 layers on natural hazards (e.g., hurricanes, wildfires), climate variables (e.g., air temperature, precipitation), soil properties, crop and land-cover types, demographics, and human health, various place and region identifiers, among other themes. These have been leveraged through the graph by a variety of applications to address challenges in food security and agricultural supply chains; sustainability related to soil conservation practices and farm labor; and delivery of emergency humanitarian aid following a disaster. In this paper, we introduce the ontology that acts as the schema for KnowWhereGraph. This broad overview provides insight into the requirements and design specifications for the graph and its schema, including the development methodology (modular ontology modeling) and the resources utilized to implement, materialize, and deploy KnowWhereGraph with its end-user interfaces and public query SPARQL endpoint.

Through end-to-end training to predict the next token, LLMs have become valuable tools for various tasks. Enhancing their core training in language modeling can improve numerous downstream applications. A successful approach to enhance language modeling uses a separate planning module to predict abstract labels of future sentences and conditions the LM on these predictions. However, this method is non-differentiable, preventing joint end-to-end tuning of the planner with the LM. We propose an effective method to improve this approach by enabling joint fine-tuning of the planner and the LM. We show that a naive way of approximating the gradient of selecting a label via the straight-through estimator is not effective. Instead, we propose to use the predicted label probabilities as mixing weights to condition the LM on a weighted average of label embeddings in a differentiable manner. This not only enables joint fine-tuning of the planner and the LM, but also allows the LM to draw on the full label distribution predicted by the planner, retaining more information. Our experimental results show consistent improvements in perplexity., Comment: 14 pages

The nonlinear damping characteristics of the oscillating wave surge converter (OWSC) significantly impact the performance of the power take-off system. This study presents a framework by integrating deep reinforcement learning (DRL) with numerical simulations of OWSC to identify optimal adaptive damping policy under varying wave conditions, thereby enhancing wave energy harvesting efficiency. Firstly, the open-source multiphysics libraries SPHinXsys and Simbody are employed to establish the numerical environment for wave interaction with OWSCs. Subsequently, a comparative analysis of three DRL algorithms-proximal policy optimization (PPO), twin delayed deep deterministic policy gradient (TD3), and soft actor-critic (SAC)-is conducted using the two-dimensional (2D) numerical study of OWSC interacting with regular waves. The results reveal that artificial neural networks capture the nonlinear characteristics of wave-structure interactions and provide efficient PTO policies. Notably, the SAC algorithm demonstrates exceptional robustness and accuracy, achieving a 10.61% improvement in wave energy harvesting. Furthermore, policies trained in a 2D environment are successfully applied to the three-dimensional (3D) study, with an improvement of 22.54% in energy harvesting. Additionally, the study shows that energy harvesting is improved by 6.42% for complex irregular waves. However, for the complex dual OWSC system, optimizing the damping characteristics alone is insufficient to enhance energy harvesting., Comment: 67 pages and 25 figures

Coherence is the most fundamental quantum resource in quantum information processing. How fast a physical system gets coherence or decoherence is a critical ingredient. We present an attainable quantum speed limit based on the variation of quantum coherence subject to a dynamical process. It indicates that for a 2-dimensional quantum state, one can always find corresponding dynamics driving it to evolve along the geodesic to another state with certain coherence variation. As applications, we study the coherence quantum speed limits of the dephasing and dissipative dynamics. It is shown that the dephasing dynamics can saturate our coherence quantum speed limit, and the decoherence of the state with identical populations will be faster than others. However, the dissipative dynamics have the opposite behavior. In addition, we illustrate a stronger tightness of our bound for the mentioned dynamics by comparison., Comment: to be published in PRA

Language models are extensively evaluated, but correctly interpreting evaluation results requires knowledge of train-test overlap which refers to the extent to which the language model is trained on the very data it is being tested on. The public currently lacks adequate information about train-test overlap: most models have no public train-test overlap statistics, and third parties cannot directly measure train-test overlap since they do not have access to the training data. To make this clear, we document the practices of 30 model developers, finding that just 9 developers report train-test overlap: 4 developers release training data under open-source licenses, enabling the community to directly measure train-test overlap, and 5 developers publish their train-test overlap methodology and statistics. By engaging with language model developers, we provide novel information about train-test overlap for three additional developers. Overall, we take the position that language model developers should publish train-test overlap statistics and/or training data whenever they report evaluation results on public test sets. We hope our work increases transparency into train-test overlap to increase the community-wide trust in model evaluations., Comment: 18 pages

We address the problem of offline learning a policy that avoids undesirable demonstrations. Unlike conventional offline imitation learning approaches that aim to imitate expert or near-optimal demonstrations, our setting involves avoiding undesirable behavior (specified using undesirable demonstrations). To tackle this problem, unlike standard imitation learning where the aim is to minimize the distance between learning policy and expert demonstrations, we formulate the learning task as maximizing a statistical distance, in the space of state-action stationary distributions, between the learning policy and the undesirable policy. This significantly different approach results in a novel training objective that necessitates a new algorithm to address it. Our algorithm, UNIQ, tackles these challenges by building on the inverse Q-learning framework, framing the learning problem as a cooperative (non-adversarial) task. We then demonstrate how to efficiently leverage unlabeled data for practical training. Our method is evaluated on standard benchmark environments, where it consistently outperforms state-of-the-art baselines. The code implementation can be accessed at: https://github.com/hmhuy0/UNIQ.

Determining the inner structure of the molecular torus around an active galactic nucleus is essential for understanding its formation mechanism. However, spatially resolving the torus is difficult because of its small size. To probe the clump conditions in the torus, we therefore perform the systematic velocity-decomposition analyses of the gaseous CO rovibrational absorption lines ($v=0\to 1,\Delta J=\pm 1$) at $\lambda\sim 4.67 \mathrm{\mu{m}}$ observed toward four (ultra)luminous infrared galaxies using the high-resolution ($R\sim 5000\text{--}10000$) spectroscopy from the Subaru Telescope. We find that each transition has two to five distinct velocity components with different line-of-sight (LOS) velocities ($V_\mathrm{LOS}\sim -240\text{--}+100\mathrm{km\,s^{-1}}$) and dispersions ($\sigma_V\sim 15\text{--}190\mathrm{km\,s^{-1}}$); i.e., the components (a), (b), ..., beginning with the broadest one in each target, indicating that the tori have clumpy structures. By assuming a hydrostatic disk ($\sigma_V\propto R_\mathrm{rot}^{-0.5}$), we find that the tori have dynamic inner structures, with the innermost component (a) outflowing with velocity $|V_\mathrm{LOS}|\sim 160\text{--}240\mathrm{km\,s^{-1}}$, and the outer components (b) and (c) outflowing more slowly or infalling with $|V_\mathrm{LOS}|\lesssim 100\mathrm{km\,s^{-1}}$. In addition, we find that the innermost component (a) can be attributed to collisionally excited hot ($\gtrsim 530$K) and dense ($n_\mathrm{H_2}\gtrsim 10^6\mathrm{cm^{-3}}$) clumps, based on the level populations. Conversely, the outer component (b) can be attributed to cold ($\sim 30\text{--}140$K) clumps radiatively excited by a far-infrared-to-submillimeter background with a brightness temperature higher than $\sim 20\text{--}400$K. These observational results demonstrate the clumpy and dynamic structure of tori in the presence of background radiation., Comment: 43 pages, 15 figures, accepted for publication in ApJ

Current benchmarks for assessing vision-language models (VLMs) often focus on their perception or problem-solving capabilities and neglect other critical aspects such as fairness, multilinguality, or toxicity. Furthermore, they differ in their evaluation procedures and the scope of the evaluation, making it difficult to compare models. To address these issues, we extend the HELM framework to VLMs to present the Holistic Evaluation of Vision Language Models (VHELM). VHELM aggregates various datasets to cover one or more of the 9 aspects: visual perception, knowledge, reasoning, bias, fairness, multilinguality, robustness, toxicity, and safety. In doing so, we produce a comprehensive, multi-dimensional view of the capabilities of the VLMs across these important factors. In addition, we standardize the standard inference parameters, methods of prompting, and evaluation metrics to enable fair comparisons across models. Our framework is designed to be lightweight and automatic so that evaluation runs are cheap and fast. Our initial run evaluates 22 VLMs on 21 existing datasets to provide a holistic snapshot of the models. We uncover new key findings, such as the fact that efficiency-focused models (e.g., Claude 3 Haiku or Gemini 1.5 Flash) perform significantly worse than their full models (e.g., Claude 3 Opus or Gemini 1.5 Pro) on the bias benchmark but not when evaluated on the other aspects. For transparency, we release the raw model generations and complete results on our website (https://crfm.stanford.edu/helm/vhelm/v2.0.1). VHELM is intended to be a living benchmark, and we hope to continue adding new datasets and models over time., Comment: NeurIPS 2024. First three authors contributed equally

In recent years, learned image compression (LIC) technologies have surpassed conventional methods notably in terms of rate-distortion (RD) performance. Most present learned techniques are VAE-based with an autoregressive entropy model, which obviously promotes the RD performance by utilizing the decoded causal context. However, extant methods are highly dependent on the fixed hand-crafted causal context. The question of how to guide the auto-encoder to generate a more effective causal context benefit for the autoregressive entropy models is worth exploring. In this paper, we make the first attempt in investigating the way to explicitly adjust the causal context with our proposed Causal Context Adjustment loss (CCA-loss). By imposing the CCA-loss, we enable the neural network to spontaneously adjust important information into the early stage of the autoregressive entropy model. Furthermore, as transformer technology develops remarkably, variants of which have been adopted by many state-of-the-art (SOTA) LIC techniques. The existing computing devices have not adapted the calculation of the attention mechanism well, which leads to a burden on computation quantity and inference latency. To overcome it, we establish a convolutional neural network (CNN) image compression model and adopt the unevenly channel-wise grouped strategy for high efficiency. Ultimately, the proposed CNN-based LIC network trained with our Causal Context Adjustment loss attains a great trade-off between inference latency and rate-distortion performance., Comment: Accepted to NeurIPS 2024

The assumption of Gaussian or Gaussian mixture data has been extensively exploited in a long series of precise performance analyses of machine learning (ML) methods, on large datasets having comparably numerous samples and features. To relax this restrictive assumption, subsequent efforts have been devoted to establish "Gaussian equivalent principles" by studying scenarios of Gaussian universality where the asymptotic performance of ML methods on non-Gaussian data remains unchanged when replaced with Gaussian data having the same mean and covariance. Beyond the realm of Gaussian universality, there are few exact results on how the data distribution affects the learning performance. In this article, we provide a precise high-dimensional characterization of empirical risk minimization, for classification under a general mixture data setting of linear factor models that extends Gaussian mixtures. The Gaussian universality is shown to break down under this setting, in the sense that the asymptotic learning performance depends on the data distribution beyond the class means and covariances. To clarify the limitations of Gaussian universality in classification of mixture data and to understand the impact of its breakdown, we specify conditions for Gaussian universality and discuss their implications for the choice of loss function., Comment: 30 pages, 3 figures

View synthesis using Neural Radiance Fields (NeRF) and Gaussian Splatting (GS) has demonstrated impressive fidelity in rendering real-world scenarios. However, practical methods for accurate and efficient epistemic Uncertainty Quantification (UQ) in view synthesis are lacking. Existing approaches for NeRF either introduce significant computational overhead (e.g., ``10x increase in training time" or ``10x repeated training") or are limited to specific uncertainty conditions or models. Notably, GS models lack any systematic approach for comprehensive epistemic UQ. This capability is crucial for improving the robustness and scalability of neural view synthesis, enabling active model updates, error estimation, and scalable ensemble modeling based on uncertainty. In this paper, we revisit NeRF and GS-based methods from a function approximation perspective, identifying key differences and connections in 3D representation learning. Building on these insights, we introduce PH-Dropout (Post hoc Dropout), the first real-time and accurate method for epistemic uncertainty estimation that operates directly on pre-trained NeRF and GS models. Extensive evaluations validate our theoretical findings and demonstrate the effectiveness of PH-Dropout., Comment: 21 pages, in submision

We consider the Kuramoto-Sakaguchi-Fokker-Planck equation (namely, parabolic Kuramoto-Sakaguchi, or Kuramoto-Sakaguchi equation, which is a nonlinear parabolic integro-differential equation) with inertia and white noise effects. We study the hydrodynamic limit of this Kuramoto-Sakaguchi equation. During showing this main result, as a support, we also prove a Hardy-type inequality over the whole real line., Comment: 34 pages

Experimentally measured characteristics of a kHz laser-driven Cu plasma X-ray source that was recently commissioned at ELI Beamlines facility are reported. The source can be driven either by an in-house developed high contrast sub-20 fs near-infrared TW laser based on optical parametric chirped-pulse amplification technology, or by a more conventional Ti:sapphire laser delivering 12 mJ, 45 fs pulses. The X-ray source parameters obtained with the two driving lasers are compared. Measured photon flux of the order up to 10^{12} K{\alpha} photons/4{\pi}/s is reported. Furthermore, experimental platforms for ultrafast X-ray diffraction and X-ray absorption and/or emission spectroscopy based on the reported source are described.

To reveal details of the internal structure, the relationship between chromospheric activity and the Rossby number has been extensively examined for main-sequence stars. For active pre-main sequence (PMS) stars, it is suggested that the level of activity be assessed using optically thin emission lines, such as Mg I. We aim to detect Mg I chromospheric emission lines from PMS stars and determine whether the chromosphere is activated by the dynamo process or by mass accretion from protoplanetary disks. We analyzed high-resolution optical spectra of $64\ $PMS stars obtained with Very Large Telescope (VLT)/X-shooter and UVES and examined the infrared Ca II (8542 A) and Mg I (8807 A) emission lines. To detect the weak chromospheric emission lines, we determined the atmospheric parameters ($T_{\rm eff}$ and $\log\ g$) and the degree of veiling of the PMS stars by comparing the observed spectra with photospheric model spectra. After subtracting the photospheric model spectrum from the PMS spectrum, we detected Ca II and Mg I as emission lines. The strengths of the Mg I emission lines in PMS stars with no veiling are comparable to those in zero-age main-sequence (ZAMS) stars if both types of stars have similar Rossby numbers. The Mg I emission lines in these PMS stars are thought to be formed by a dynamo process similar to that in ZAMS stars. In contrast, the Mg I emission lines in PMS stars with veiling are stronger than those in ZAMS stars. These objects are believed to have protoplanetary disks, where mass accretion generates shocks near the photosphere, heating the chromosphere. The chromosphere of PMS stars is activated not only by the dynamo process but also by mass accretion., Comment: accepted for Astronomy and Astrophysics

Offline reinforcement learning (RL) has garnered significant attention for its ability to learn effective policies from pre-collected datasets without the need for further environmental interactions. While promising results have been demonstrated in single-agent settings, offline multi-agent reinforcement learning (MARL) presents additional challenges due to the large joint state-action space and the complexity of multi-agent behaviors. A key issue in offline RL is the distributional shift, which arises when the target policy being optimized deviates from the behavior policy that generated the data. This problem is exacerbated in MARL due to the interdependence between agents' local policies and the expansive joint state-action space. Prior approaches have primarily addressed this challenge by incorporating regularization in the space of either Q-functions or policies. In this work, we introduce a regularizer in the space of stationary distributions to better handle distributional shift. Our algorithm, ComaDICE, offers a principled framework for offline cooperative MARL by incorporating stationary distribution regularization for the global learning policy, complemented by a carefully structured multi-agent value decomposition strategy to facilitate multi-agent training. Through extensive experiments on the multi-agent MuJoCo and StarCraft II benchmarks, we demonstrate that ComaDICE achieves superior performance compared to state-of-the-art offline MARL methods across nearly all tasks.

We present Exact Volumetric Ellipsoid Rendering (EVER), a method for real-time differentiable emission-only volume rendering. Unlike recent rasterization based approach by 3D Gaussian Splatting (3DGS), our primitive based representation allows for exact volume rendering, rather than alpha compositing 3D Gaussian billboards. As such, unlike 3DGS our formulation does not suffer from popping artifacts and view dependent density, but still achieves frame rates of $\sim\!30$ FPS at 720p on an NVIDIA RTX4090. Since our approach is built upon ray tracing it enables effects such as defocus blur and camera distortion (e.g. such as from fisheye cameras), which are difficult to achieve by rasterization. We show that our method is more accurate with fewer blending issues than 3DGS and follow-up work on view-consistent rendering, especially on the challenging large-scale scenes from the Zip-NeRF dataset where it achieves sharpest results among real-time techniques., Comment: Project page: https://half-potato.gitlab.io/posts/ever

Fluid-structure interaction (FSI) problems are characterized by strong nonlinearities arising from complex interactions between fluids and structures. These pose significant challenges for traditional control strategies in optimizing structural motion, often leading to suboptimal performance. In contrast, deep reinforcement learning (DRL), through agent interactions within numerical simulation environments and the approximation of control policies using deep neural networks (DNNs), has shown considerable promise in addressing high-dimensional FSI problems. Additionally, smoothed particle hydrodynamics (SPH) offers a flexible and efficient computational approach for modeling large deformations, fractures, and complex interface movements inherent in FSI, outperforming traditional grid-based methods. In this work, we present DRLinSPH, an open-source Python platform that integrates the SPH-based numerical environment provided by the open-source software SPHinXsys with the mature DRL platform Tianshou to enable parallel training for FSI problems. DRLinSPH has been successfully applied to four FSI scenarios: sloshing suppression using rigid and elastic baffles, optimization of wave energy capture through an oscillating wave surge converter (OWSC), and muscle-driven fish swimming in vortices. The results demonstrate the platform's accuracy, stability, and scalability, highlighting its potential to advance industrial solutions for complex FSI challenges., Comment: 68 pages 31 figures

Steering models (such as the generalized two-point model) predict human steering behavior well when the human is in direct control of a vehicle. In vehicles under autonomous control, human control inputs are not used; rather, an autonomous controller applies steering and acceleration commands to the vehicle. For example, human steering input may be used for state estimation rather than direct control. We show that human steering behavior changes when the human no longer directly controls the vehicle and the two are instead working in a shared autonomy paradigm. Thus, when a vehicle is not under direct human control, steering models like the generalized two-point model do not predict human steering behavior. We also show that the error between predicted human steering behavior and actual human steering behavior reflects a fundamental difference when the human directly controls the vehicle compared to when the vehicle is autonomously controlled. Moreover, we show that a single distribution describes the error between predicted human steering behavior and actual human steering behavior when the human's steering inputs are used for state estimation and the vehicle is autonomously controlled, indicating there may be a underlying model for human steering behavior under this type of shared autonomous control. Future work includes determining this shared autonomous human steering model and demonstrating its performance., Comment: 6 pages, 10 figures, accepted for publication at the 5th IFAC at the 5th IFAC Workshop on Cyber-Physical Human Systems

Harnessing quantum interference among bosons provides significant opportunities as bosons often carry longer coherence time than fermions. As an example of quantum interference, Fano resonance involving phonons or photons describes the coupling between discrete and continuous states, signified by an asymmetric spectral lineshape. Utilizing photon-based Fano resonance, molecule sensing with ultra-high sensitivity and ultrafast optical switching has been realized. However, phonon-based Fano resonance, which would expand the application space to a vaster regime, has been less exploited because of the weak coupling between discrete phonons with continuous states such as electronic continuum. In this work, we report the discovery of giant phonon-based Fano resonance in a graphene/2D Ag/SiC heterostructure. The Fano asymmetry, being proportional to the coupling strength, exceeds prior reports by two orders of magnitude. This Fano asymmetry arises from simultaneous frequency and lifetime matching between discrete and continuous phonons of SiC. The introduction of 2D Ag layers restructures SiC at the interface and facilitates resonant scattering to further enhance the Fano asymmetry, which is not achievable with conventional Ag thin films. With these unique properties, we demonstrated that the phonon-based Fano resonance can be used for ultrasensitive molecule detection at the single-molecule level. Our work highlights strong Fano resonance in the phononic system, opening avenues for engineering quantum interference based on bosons. Further, our findings provide opportunities for advancing phonon-related applications, including biochemical sensing, quantum transduction, and superconductor-based quantum computing.

Parameter-efficient transfer learning (PETL) has attracted significant attention lately, due to the increasing size of pre-trained models and the need to fine-tune (FT) them for superior downstream performance. This community-wide enthusiasm has sparked a plethora of approaches. Nevertheless, a systematic study to understand their performance and suitable application scenarios is lacking, leaving questions like when to apply PETL and which approach to use largely unanswered. In this paper, we conduct a unifying empirical study of representative PETL methods in the context of Vision Transformers. We systematically tune their hyper-parameters to fairly compare their accuracy on downstream tasks. Our study not only offers a valuable user guide but also unveils several new insights. First, if tuned carefully, different PETL methods can obtain similar accuracy in the low-shot benchmark VTAB-1K. This includes simple methods like FT the bias terms that were reported inferior. Second, though with similar accuracy, we find that PETL methods make different mistakes and high-confidence predictions, likely due to their different inductive biases. Such an inconsistency (or complementariness) opens up the opportunity for ensemble methods, and we make preliminary attempts at this. Third, going beyond the commonly used low-shot tasks, we find that PETL is also useful in many-shot regimes -- it achieves comparable and sometimes better accuracy than full FT, using much fewer learnable parameters. Last but not least, we investigate PETL's ability to preserve a pre-trained model's robustness to distribution shifts (e.g., a CLIP backbone). Perhaps not surprisingly, PETL methods outperform full FT alone. However, with weight-space ensembles, the fully fine-tuned model can better balance target (i.e., downstream) distribution and distribution shift performance, suggesting a future research direction for PETL., Comment: Code is available at https://github.com/OSU-MLB/PETL_Vision

Fine-tuning is arguably the most straightforward way to tailor a pre-trained model (e.g., a foundation model) to downstream applications, but it also comes with the risk of losing valuable knowledge the model had learned in pre-training. For example, fine-tuning a pre-trained classifier capable of recognizing a large number of classes to master a subset of classes at hand is shown to drastically degrade the model's accuracy in the other classes it had previously learned. As such, it is hard to further use the fine-tuned model when it encounters classes beyond the fine-tuning data. In this paper, we systematically dissect the issue, aiming to answer the fundamental question, "What has been damaged in the fine-tuned model?" To our surprise, we find that the fine-tuned model neither forgets the relationship among the other classes nor degrades the features to recognize these classes. Instead, the fine-tuned model often produces more discriminative features for these other classes, even if they were missing during fine-tuning! {What really hurts the accuracy is the discrepant logit scales between the fine-tuning classes and the other classes}, implying that a simple post-processing calibration would bring back the pre-trained model's capability and at the same time unveil the feature improvement over all classes. We conduct an extensive empirical study to demonstrate the robustness of our findings and provide preliminary explanations underlying them, suggesting new directions for future theoretical analysis. Our code is available at https://github.com/OSU-MLB/Fine-Tuning-Is-Fine-If-Calibrated., Comment: The paper has been accepted to NeurIPS 2024. The first three authors contribute equally

Societal risk emanating from how recommender algorithms disseminate content online is now well documented. Emergent regulation aims to mitigate this risk through ethical audits and enabling new research on the social impact of algorithms. However, there is currently a need for tools and methods that enable such evaluation. This paper presents ARTAI, an evaluation environment that enables large-scale assessments of recommender algorithms to identify harmful patterns in how content is distributed online and enables the implementation of new regulatory requirements for increased transparency in recommender systems., Comment: 3 pages, 1 figure, accepted at FAccTRec 2024 Workshop, RecSys 2024

In tackling the challenge of Multi-Document Summarization (MDS), numerous methods have been proposed, spanning both extractive and abstractive summarization techniques. However, each approach has its own limitations, making it less effective to rely solely on either one. An emerging and promising strategy involves a synergistic fusion of extractive and abstractive summarization methods. Despite the plethora of studies in this domain, research on the combined methodology remains scarce, particularly in the context of Vietnamese language processing. This paper presents a novel Vietnamese MDS framework leveraging a two-component pipeline architecture that integrates extractive and abstractive techniques. The first component employs an extractive approach to identify key sentences within each document. This is achieved by a modification of the pre-trained BERT network, which derives semantically meaningful phrase embeddings using siamese and triplet network structures. The second component utilizes the VBD-LLaMA2-7B-50b model for abstractive summarization, ultimately generating the final summary document. Our proposed framework demonstrates a positive performance, attaining ROUGE-2 scores of 39.6% on the VN-MDS dataset and outperforming the state-of-the-art baselines., Comment: 10 pages

This paper presents a diffusion-based recommender system that incorporates classifier-free guidance. Most current recommender systems provide recommendations using conventional methods such as collaborative or content-based filtering. Diffusion is a new approach to generative AI that improves on previous generative AI approaches such as Variational Autoencoders (VAEs) and Generative Adversarial Networks (GANs). We incorporate diffusion in a recommender system that mirrors the sequence users take when browsing and rating items. Although a few current recommender systems incorporate diffusion, they do not incorporate classifier-free guidance, a new innovation in diffusion models as a whole. In this paper, we present a diffusion recommender system that augments the underlying recommender system model for improved performance and also incorporates classifier-free guidance. Our findings show improvements over state-of-the-art recommender systems for most metrics for several recommendation tasks on a variety of datasets. In particular, our approach demonstrates the potential to provide better recommendations when data is sparse., Comment: 8 pages

In this work, we address the objective of protecting the states of a distributed dynamical system from eavesdropping adversaries. We prove that state-of-the-art distributed algorithms, which rely on communicating the agents' states, are vulnerable in that the final states can be perfectly estimated by any adversary including those with arbitrarily small eavesdropping success probability. While existing literature typically adds an extra layer of protection, such as encryption or differential privacy techniques, we demonstrate the emergence of a fundamental protection quotient in distributed systems when innovation signals are communicated instead of the agents' states.

To assess prior mean-field studies that the interacting Kane-Mele model supports a novel antiferromagnetic Chern insulating phase (AFCI) for a wide range of sublattice potentials, we analyze the Kane-Mele-Hubbard model in the presence of a sublattice potential using determinant quantum Monte Carlo (DQMC). Instead of an AFCI, we find that the ground state is a quantum spin Hall (QSH) insulator for intermediate values of the sublattice potential $\lambda_v$, albeit with a small gap. The QSH state gives way to a trivial band insulator as the sublattice potential increases beyond a critical value. Only at small sublattice potentials does the QSH state transition into a trivial Mott insulator with xy antiferromagnetic correlations. The QAH feature is only observed at high temperature. The QAH feature crosses over to an incipient QSH state when the topological gap stabilizes. Our work is consistent with the experimental observation that in twisted bilayer MoTe$_2$ and WSe$_2$ as well as AB stacked MoTe$_2$/WSe$_2$, where QSH is consistently observed at even-integer filling over a wide range of parameters.

Extending geodetic and astrometric Very Long Baseline Interferometry (VLBI) observations from traditional centimeter wavebands to millimeter wavebands offers numerous scientific potentials and benefits. However, it was considered quite challenging due to various factors, including the increased effects of atmospheric opacity and turbulence at millimeter wavelengths. Here, we present the results of the first geodetic-mode VLBI experiment, simultaneously observing 82 sources at 22/43/88/132 GHz (K/Q/W/D bands) using the Korean VLBI Network (KVN). We introduced the frequency phase transfer (FPT) method to geodetic VLBI analysis, an approach for calibrating atmospheric phase fluctuations at higher frequencies by transferring phase solutions from lower frequencies. With a 2-minute scan, FPT improved the signal-to-noise ratio (SNR) of most fringes, some by over 100%, thereby enhancing the detection rate of weak sources at millimeter wavebands. Additionally, FPT reduced systematic errors in group delay and delay rate, with the weighted root-mean-squares (WRMS) of the post-fitting residuals decreasing from 25.0 ps to 20.5 ps at the W band and from 39.3 ps to 27.6 ps at the D band. There were no notable differences observed in calibrating atmospheric phase fluctuations at the K band (WRMS = 12.4 ps) and Q band (WRMS = 11.8 ps) with the KVN baselines. This experiment demonstrated that the millimeter waveband can be used for geodetic and astrometric applications with high precision., Comment: 17 pages, 6 figures, accepted by The Astronomical Journal. Your feedback to improve the manuscript would be greatly appreciated

The task of audio-driven portrait animation involves generating a talking head video using an identity image and an audio track of speech. While many existing approaches focus on lip synchronization and video quality, few tackle the challenge of generating emotion-driven talking head videos. The ability to control and edit emotions is essential for producing expressive and realistic animations. In response to this challenge, we propose EMOdiffhead, a novel method for emotional talking head video generation that not only enables fine-grained control of emotion categories and intensities but also enables one-shot generation. Given the FLAME 3D model's linearity in expression modeling, we utilize the DECA method to extract expression vectors, that are combined with audio to guide a diffusion model in generating videos with precise lip synchronization and rich emotional expressiveness. This approach not only enables the learning of rich facial information from emotion-irrelevant data but also facilitates the generation of emotional videos. It effectively overcomes the limitations of emotional data, such as the lack of diversity in facial and background information, and addresses the absence of emotional details in emotion-irrelevant data. Extensive experiments and user studies demonstrate that our approach achieves state-of-the-art performance compared to other emotion portrait animation methods., Comment: 12 pages, 7 figures

Quantile regression, a robust method for estimating conditional quantiles, has advanced significantly in fields such as econometrics, statistics, and machine learning. In high-dimensional settings, where the number of covariates exceeds sample size, penalized methods like lasso have been developed to address sparsity challenges. Bayesian methods, initially connected to quantile regression via the asymmetric Laplace likelihood, have also evolved, though issues with posterior variance have led to new approaches, including pseudo/score likelihoods. This paper presents a novel probabilistic machine learning approach for high-dimensional quantile prediction. It uses a pseudo-Bayesian framework with a scaled Student-t prior and Langevin Monte Carlo for efficient computation. The method demonstrates strong theoretical guarantees, through PAC-Bayes bounds, that establish non-asymptotic oracle inequalities, showing minimax-optimal prediction error and adaptability to unknown sparsity. Its effectiveness is validated through simulations and real-world data, where it performs competitively against established frequentist and Bayesian techniques.

Model Inversion (MI) is a type of privacy violation that focuses on reconstructing private training data through abusive exploitation of machine learning models. To defend against MI attacks, state-of-the-art (SOTA) MI defense methods rely on regularizations that conflict with the training loss, creating explicit tension between privacy protection and model utility. In this paper, we present a new method to defend against MI attacks. Our method takes a new perspective and focuses on training data. Our idea is based on a novel insight on Random Erasing (RE), which has been applied in the past as a data augmentation technique to improve the model accuracy under occlusion. In our work, we instead focus on applying RE for degrading MI attack accuracy. Our key insight is that MI attacks require significant amount of private training data information encoded inside the model in order to reconstruct high-dimensional private images. Therefore, we propose to apply RE to reduce private information presented to the model during training. We show that this can lead to substantial degradation in MI reconstruction quality and attack accuracy. Meanwhile, natural accuracy of the model is only moderately affected. Our method is very simple to implement and complementary to existing defense methods. Our extensive experiments of 23 setups demonstrate that our method can achieve SOTA performance in balancing privacy and utility of the models. The results consistently demonstrate the superiority of our method over existing defenses across different MI attacks, network architectures, and attack configurations., Comment: Under review. The first two authors contributed equally

Stereoelectroencephalography (SEEG) is an invasive surgical procedure to record the electrical activities in cortical brain regions, aiming at identifying the Epileptogenic Zone (EZ) in patients with drug-resistant epilepsy. To improve the accuracy of the EZ definition, SEEG analysis can be supported by computational tools, among which the Epileptogenic Index (EI) represents the most common solution. However, the scientific community has still not found an agreement on which quantitative biomarkers can characterize the cortical sites within the EZ. In this work, we design a new algorithm, named Desynchronization Index (DI), to assist neurophysiologists in SEEG interpretation. Our algorithm estimates the effective connectivity between cortical sites and hypothesizes that the EZ is identified by those sites getting abnormally desynchronized from the network during the seizure generation. We test the proposed method over a SEEG dataset of 10 seizures, comparing its accuracy in terms of EZ definition against the EI algorithm and clinical ground truth. Our results indicate that the DI algorithm underscores specific connectivity dynamics that can hardly be identified with a pure visual analysis, increasing sensitivity in detecting epileptogenic cortical sites.

In this paper, we assess the movement error of a targeting system given target location data from artificial intelligence (AI) methods in automatic target recognition (ATR) systems. Few studies evaluate the impacts on the accuracy in moving a targeting system to an aimpoint provided in this manner. To address this knowledge gap, we assess the performance of a controlled gun turret system given target location from an object detector developed from AI methods. In our assessment, we define a measure of object detector error and examine the correlations with several standard metrics in object detection. We then statistically analyze the object detector error data and turret movement error data acquired from controlled targeting simulations, as well as their aggregate error, to examine the impact on turret movement accuracy. Finally, we study the correlations between additional metrics and the probability of a hit. The results indicate that AI technologies are a significant source of error to targeting systems. Moreover, the results suggest that metrics such as the confidence score, intersection-over-union, average precision and average recall are predictors of accuracy against stationary targets with our system parameters., Comment: 30 pages, 21 figures

The purpose of this paper is to characterize aiming errors in controlled weapon systems given target location as input. To achieve this objective, we analyze the accuracy of a controlled weapon system model for stationary and moving targets under different error sources and firing times. First, we develop a mathematical model of a gun turret and use it to design two controllers, a Proportional-Integral-Derivative controller and a Model Predictive controller, which accept the target location input and move the turret to the centroid of the target in simulations. For stationary targets, we analyze the impact of errors in estimating the system's parameters and uncertainty in the aim point measurement. Our results indicate that turret movement is more sensitive to errors in the moment of inertia than the damping coefficient, which could lead to incorrect simulations of controlled turret system accuracy. The results also support the hypothesis that turret movement errors are larger over longer distances of gun turret movement and, assuming no time constraints, accuracy improves the longer one waits to fire; though this may not always be practical in a combat scenario. Additionally, we demonstrate that the integral control component is needed for high accuracy in moving target scenarios., Comment: 29 pages, 15 figures

Facial expression recognition (FER) aims to analyze emotional states from static images and dynamic sequences, which is pivotal in enhancing anthropomorphic communication among humans, robots, and digital avatars by leveraging AI technologies. As the FER field evolves from controlled laboratory environments to more complex in-the-wild scenarios, advanced methods have been rapidly developed and new challenges and apporaches are encounted, which are not well addressed in existing reviews of FER. This paper offers a comprehensive survey of both image-based static FER (SFER) and video-based dynamic FER (DFER) methods, analyzing from model-oriented development to challenge-focused categorization. We begin with a critical comparison of recent reviews, an introduction to common datasets and evaluation criteria, and an in-depth workflow on FER to establish a robust research foundation. We then systematically review representative approaches addressing eight main challenges in SFER (such as expression disturbance, uncertainties, compound emotions, and cross-domain inconsistency) as well as seven main challenges in DFER (such as key frame sampling, expression intensity variations, and cross-modal alignment). Additionally, we analyze recent advancements, benchmark performances, major applications, and ethical considerations. Finally, we propose five promising future directions and development trends to guide ongoing research. The project page for this paper can be found at https://github.com/wangyanckxx/SurveyFER.

Multimodal sentiment analysis aims to effectively integrate information from various sources to infer sentiment, where in many cases there are no annotations for unimodal labels. Therefore, most works rely on multimodal labels for training. However, there exists the noisy label problem for the learning of unimodal signals as multimodal annotations are not always the ideal substitutes for the unimodal ones, failing to achieve finer optimization for individual modalities. In this paper, we explore the learning of unimodal labels under the weak supervision from the annotated multimodal labels. Specifically, we propose a novel meta uni-label generation (MUG) framework to address the above problem, which leverages the available multimodal labels to learn the corresponding unimodal labels by the meta uni-label correction network (MUCN). We first design a contrastive-based projection module to bridge the gap between unimodal and multimodal representations, so as to use multimodal annotations to guide the learning of MUCN. Afterwards, we propose unimodal and multimodal denoising tasks to train MUCN with explicit supervision via a bi-level optimization strategy. We then jointly train unimodal and multimodal learning tasks to extract discriminative unimodal features for multimodal inference. Experimental results suggest that MUG outperforms competitive baselines and can learn accurate unimodal labels.

There has been much discussion about the initial conditions of the early Universe in the context of quantum theory. In this paper, we construct the wave function and probability distribution by adopting the quantum version of the BPS equation instead of the usual Wheeler-DeWitt equation in a minisuperspace quantum cosmology with spatially uniform scalar fields. Although the model treated in this study is technically valid for a limited form of scalar potential, we show that it is possible to construct a conserved probability current in our model. We also examine classical and quantum aspects of models with the Dirac-Born-Infeld type scalar fields., Comment: 19 pages, 1 figure

Modern language models predict the next token in the sequence by considering the past text through a powerful function such as attention. However, language models have no explicit mechanism that allows them to spend computation time for planning long-distance future text, leading to a suboptimal token prediction. In this paper, we propose a planner that predicts a latent plan for many sentences into the future. By sampling multiple plans at once, we condition the language model on an accurate approximation of the distribution of text continuations, which leads to better next token prediction accuracy. In effect, this allows trading computation time for prediction accuracy., Comment: preprint

Multimodal Emotion Recognition in Conversations (ERC) is a typical multimodal learning task in exploiting various data modalities concurrently. Prior studies on effective multimodal ERC encounter challenges in addressing modality imbalances and optimizing learning across modalities. Dealing with these problems, we present a novel framework named Ada2I, which consists of two inseparable modules namely Adaptive Feature Weighting (AFW) and Adaptive Modality Weighting (AMW) for feature-level and modality-level balancing respectively via leveraging both Inter- and Intra-modal interactions. Additionally, we introduce a refined disparity ratio as part of our training optimization strategy, a simple yet effective measure to assess the overall discrepancy of the model's learning process when handling multiple modalities simultaneously. Experimental results validate the effectiveness of Ada2I with state-of-the-art performance compared to baselines on three benchmark datasets, particularly in addressing modality imbalances., Comment: Accepted at ACM Multimedia 2024

We measure the ionised gas velocity dispersions of star-forming galaxies in the MAGPI survey ($z\sim0.3$) and compare them with galaxies in the SAMI ($z\sim0.05$) and KROSS ($z\sim1$) surveys to investigate how the ionised gas velocity dispersion evolves. For the first time, we use a consistent method that forward models galaxy kinematics from $z=0$ to $z=1$. This method accounts for spatial substructure in emission line flux and beam smearing. We investigate the correlation between gas velocity dispersion and galaxy properties to understand the mechanisms that drive gas turbulence. We find that in both MAGPI and SAMI galaxies, the gas velocity dispersion more strongly correlates with the star-formation rate surface density ($\Sigma_{\rm SFR}$) than with a variety of other physical properties, and the average gas velocity dispersion is similar, at the same $\Sigma_{\rm SFR}$, for SAMI, MAGPI and KROSS galaxies. The results indicate that mechanisms related to $\Sigma_{\rm SFR}$ could be the dominant driver of gas turbulence from $z\sim1$ to $z\sim0$, for example, stellar feedback and/or gravitational instability. The gas velocity dispersion of MAGPI galaxies is also correlated with the non-rotational motion of the gas, illustrating that in addition to star-formation feedback, gas transportation and accretion may also contribute to the gas velocity dispersion for galaxies at $z\sim 0.3$. KROSS galaxies only have a moderate correlation between gas velocity dispersion and $\Sigma_{\rm SFR}$ and a higher scatter of gas velocity dispersion with respect to $\Sigma_{\rm SFR}$, in agreement with the suggestion that other mechanisms, such as gas transportation and accretion, are relatively more important at higher redshift galaxies., Comment: 15 pages, 10 figures, accepted for publication in MNRAS

The emergence of Large Language Models (LLMs) and multimodal foundation models (FMs) has generated heightened interest in their applications that integrate vision and language. This paper investigates the capabilities of ChatGPT-4V and Gemini Pro for Street View Imagery, Built Environment, and Interior by evaluating their performance across various tasks. The assessments include street furniture identification, pedestrian and car counts, and road width measurement in Street View Imagery; building function classification, building age analysis, building height analysis, and building structure classification in the Built Environment; and interior room classification, interior design style analysis, interior furniture counts, and interior length measurement in Interior. The results reveal proficiency in length measurement, style analysis, question answering, and basic image understanding, but highlight limitations in detailed recognition and counting tasks. While zero-shot learning shows potential, performance varies depending on the problem domains and image complexities. This study provides new insights into the strengths and weaknesses of multimodal foundation models for practical challenges in Street View Imagery, Built Environment, and Interior. Overall, the findings demonstrate foundational multimodal intelligence, emphasizing the potential of FMs to drive forward interdisciplinary applications at the intersection of computer vision and language.

Neural radiance fields (NeRFs) generally require many images with accurate poses for accurate novel view synthesis, which does not reflect realistic setups where views can be sparse and poses can be noisy. Previous solutions for learning NeRFs with sparse views and noisy poses only consider local geometry consistency with pairs of views. Closely following \textit{bundle adjustment} in Structure-from-Motion (SfM), we introduce TrackNeRF for more globally consistent geometry reconstruction and more accurate pose optimization. TrackNeRF introduces \textit{feature tracks}, \ie connected pixel trajectories across \textit{all} visible views that correspond to the \textit{same} 3D points. By enforcing reprojection consistency among feature tracks, TrackNeRF encourages holistic 3D consistency explicitly. Through extensive experiments, TrackNeRF sets a new benchmark in noisy and sparse view reconstruction. In particular, TrackNeRF shows significant improvements over the state-of-the-art BARF and SPARF by $\sim8$ and $\sim1$ in terms of PSNR on DTU under various sparse and noisy view setups. The code is available at \href{https://tracknerf.github.io/}., Comment: ECCV 2024 (supplemental pages included)

Utilizing code snippets on Stack Overflow (SO) is a common practice among developers for problem-solving. Although SO code snippets serve as valuable resources, it is important to acknowledge their imperfections, reusing problematic code snippets can lead to the introduction of suboptimal or buggy code into software projects. SO comments often point out weaknesses of a post and provide valuable insights to improve the quality of answers, while SO comments are usually missed and/or ignored, leaving these problematic code snippets untouched. In this work, we first investigate the task of automatic SO posts updating based on their associated comments. We introduce a novel framework, named Soup (Stack Overflow Updator for Post) for this task. Soup addresses two key tasks: Valid Comment-Edit Prediction (VCP) and Automatic Post Updating (APU). Extensive experimental results show the promising performance of our model over a set of benchmarks. Moreover, we also performed an in-the-wild evaluation on Stack Overflow, we submitted 50 edits generated by our approach to Stack Overflow posts and 21 of them have been verified and accepted by SO maintainers, further proving the practical value of Soup.

PAC-Bayesian bounds have proven to be a valuable tool for deriving generalization bounds and for designing new learning algorithms in machine learning. However, it typically focus on providing generalization bounds with respect to a chosen loss function. In classification tasks, due to the non-convex nature of the 0-1 loss, a convex surrogate loss is often used, and thus current PAC-Bayesian bounds are primarily specified for this convex surrogate. This work shifts its focus to providing misclassification excess risk bounds for PAC-Bayesian classification when using a convex surrogate loss. Our key ingredient here is to leverage PAC-Bayesian relative bounds in expectation rather than relying on PAC-Bayesian bounds in probability. We demonstrate our approach in several important applications.

In the pathway toward Artificial General Intelligence (AGI), understanding human's affection is essential to enhance machine's cognition abilities. For achieving more sensual human-AI interaction, Multimodal Affective Computing (MAC) in human-spoken videos has attracted increasing attention. However, previous methods are mainly devoted to designing multimodal fusion algorithms, suffering from two issues: semantic imbalance caused by diverse pre-processing operations and semantic mismatch raised by inconsistent affection content contained in different modalities comparing with the multimodal ground truth. Besides, the usage of manual features extractors make they fail in building end-to-end pipeline for multiple MAC downstream tasks. To address above challenges, we propose a novel end-to-end framework named SemanticMAC to compute multimodal semantic-centric affection for human-spoken videos. We firstly employ pre-trained Transformer model in multimodal data pre-processing and design Affective Perceiver module to capture unimodal affective information. Moreover, we present a semantic-centric approach to unify multimodal representation learning in three ways, including gated feature interaction, multi-task pseudo label generation, and intra-/inter-sample contrastive learning. Finally, SemanticMAC effectively learn specific- and shared-semantic representations in the guidance of semantic-centric labels. Extensive experimental results demonstrate that our approach surpass the state-of-the-art methods on 7 public datasets in four MAC downstream tasks., Comment: Under Review

Nature disasters play a key role in shaping human-urban infrastructure interactions. Effective and efficient response to natural disasters is essential for building resilience and a sustainable urban environment. Two types of information are usually the most necessary and difficult to gather in disaster response. The first information is about disaster damage perception, which shows how badly people think that urban infrastructure has been damaged. The second information is geolocation awareness, which means how people whereabouts are made available. In this paper, we proposed a novel disaster mapping framework, namely CVDisaster, aiming at simultaneously addressing geolocalization and damage perception estimation using cross-view Street-View Imagery (SVI) and Very High-Resolution satellite imagery. CVDisaster consists of two cross-view models, where CVDisaster-Geoloc refers to a cross-view geolocalization model based on a contrastive learning objective with a Siamese ConvNeXt image encoder, and CVDisaster-Est is a cross-view classification model based on a Couple Global Context Vision Transformer (CGCViT). Taking Hurricane IAN as a case study, we evaluate the CVDisaster framework by creating a novel cross-view dataset (CVIAN) and conducting extensive experiments. As a result, we show that CVDisaster can achieve highly competitive performance (over 80% for geolocalization and 75% for damage perception estimation) with even limited fine-tuning efforts, which largely motivates future cross-view models and applications within a broader GeoAI research community. The data and code are publicly available at: https://github.com/tum-bgd/CVDisaster.

The accuracy of mapping agricultural fields across large areas is steadily improving with high-resolution satellite imagery and deep learning (DL) models, even in regions where fields are small and geometrically irregular. However, developing effective DL models often requires large, expensive label datasets, typically available only for specific years or locations. This limits the ability to create annual maps essential for agricultural monitoring, as domain shifts occur between years and regions due to changes in farming practices and environmental conditions. The challenge is to design a model flexible enough to account for these shifts without needing yearly labels. While domain adaptation techniques or semi-supervised training are common solutions, we explored enhancing the model's generalization power. Our results indicate that a holistic approach is essential, combining methods to improve generalization. Specifically, using an area-based loss function, such as Tversky-focal loss (TFL), significantly improved predictions across multiple years. The use of different augmentation techniques helped to encode different types of invariance, particularly photometric augmentations encoded invariance to brightness changes, though they increased false positives. The combination of photometric augmentation, TFL loss, and MC-dropout produced the best results, although dropout alone led to more false negatives in subsequent year predictions. Additionally, the choice of input normalization had a significant impact, with the best results obtained when statistics were calculated either locally or across the entire dataset over all bands (lab and gab). We developed a workflow that enabled a U-Net model to generate effective multi-year crop maps over large areas. Our code, available at: https://github.com/agroimpacts/cnn-generalization-enhancement, will be regularly updated with improvements.