Search

Large Language Models (LLMs) have revolutionized various domains but remain vulnerable to prompt injection attacks, where malicious inputs manipulate the model into ignoring original instructions and executing designated action. In this paper, we investigate the underlying mechanisms of these attacks by analyzing the attention patterns within LLMs. We introduce the concept of the distraction effect, where specific attention heads, termed important heads, shift focus from the original instruction to the injected instruction. Building on this discovery, we propose Attention Tracker, a training-free detection method that tracks attention patterns on instruction to detect prompt injection attacks without the need for additional LLM inference. Our method generalizes effectively across diverse models, datasets, and attack types, showing an AUROC improvement of up to 10.0% over existing methods, and performs well even on small LLMs. We demonstrate the robustness of our approach through extensive evaluations and provide insights into safeguarding LLM-integrated systems from prompt injection vulnerabilities., Comment: Project page: https://huggingface.co/spaces/TrustSafeAI/Attention-Tracker

State-of-the-art (SOTA) semi-supervised learning techniques, such as FixMatch and it's variants, have demonstrated impressive performance in classification tasks. However, these methods are not directly applicable to regression tasks. In this paper, we present RankUp, a simple yet effective approach that adapts existing semi-supervised classification techniques to enhance the performance of regression tasks. RankUp achieves this by converting the original regression task into a ranking problem and training it concurrently with the original regression objective. This auxiliary ranking classifier outputs a classification result, thus enabling integration with existing semi-supervised classification methods. Moreover, we introduce regression distribution alignment (RDA), a complementary technique that further enhances RankUp's performance by refining pseudo-labels through distribution alignment. Despite its simplicity, RankUp, with or without RDA, achieves SOTA results in across a range of regression benchmarks, including computer vision, audio, and natural language processing tasks. Our code and log data are open-sourced at https://github.com/pm25/semi-supervised-regression., Comment: Accepted at NeurIPS 2024 (Poster)

This study investigates the transverse momentum ($p_T$) spectra of identified light charged hadrons produced in gold--gold (Au+Au) collisions across various centrality classes at center-of-mass energies per nucleon pair, $\sqrt{s_{NN}}$, ranging from 7.7 to 200 GeV, as measured by the STAR Collaboration at the Relativistic Heavy Ion Collider (RHIC). The analysis employs standard (Bose-Einstein/Fermi-Dirac), Tsallis, and q-dual statistics to fit the same $p_T$ spectra and derive distinct effective temperatures: $T_{\text{Standard}}$, $T_{\text{Tsallis}}$, and $T_{\text{q-dual}}$. In most instances, there exists an approximately linear relationship or positive correlation between $T_{\text{Tsallis}}$ and $T_{\text{Standard}}$, as well as between $T_{\text{q-dual}}$ and $T_{\text{Standard}}$, when considering $T_{\text{Standard}}$ as a baseline. However, while both $T_{\text{Tsallis}}$ and $T_{\text{q-dual}}$ increase from semi-central to central Au+Au collisions at 62.4 GeV and 200 GeV, where QGP is expected, changes in $T_{\text{Standard}}$ occur more gradually. This work suggests that $T_{\text{Standard}}$ is better suited for characterizing phase transitions between hadronic matter and QGP compared to $T_{\text{Tsallis}}$ or $T_{\text{q-dual}}$, primarily due to the considerations related to entropy index in the Tsallis and q-dual statistics., Comment: 19 pages, 9 figures

Brown constructed a series of threefold flips given by the GIT quotient of a hypersurface in C5. In this article, we classify threefold flips and flops which are the GIT quotients of complete intersections in C6. We also show that there are no more new examples as GIT quotients of complete intersections in Cn with n >= 7., Comment: 23 pages, 1 figure, 7 tables

Laboratory accidents pose significant risks to human life and property, underscoring the importance of robust safety protocols. Despite advancements in safety training, laboratory personnel may still unknowingly engage in unsafe practices. With the increasing reliance on large language models (LLMs) for guidance in various fields, including laboratory settings, there is a growing concern about their reliability in critical safety-related decision-making. Unlike trained human researchers, LLMs lack formal lab safety education, raising questions about their ability to provide safe and accurate guidance. Existing research on LLM trustworthiness primarily focuses on issues such as ethical compliance, truthfulness, and fairness but fails to fully cover safety-critical real-world applications, like lab safety. To address this gap, we propose the Laboratory Safety Benchmark (LabSafety Bench), a comprehensive evaluation framework based on a new taxonomy aligned with Occupational Safety and Health Administration (OSHA) protocols. This benchmark includes 765 multiple-choice questions verified by human experts, assessing LLMs and vision language models (VLMs) performance in lab safety contexts. Our evaluations demonstrate that while GPT-4o outperforms human participants, it is still prone to critical errors, highlighting the risks of relying on LLMs in safety-critical environments. Our findings emphasize the need for specialized benchmarks to accurately assess the trustworthiness of LLMs in real-world safety applications., Comment: 50 pages, 19 figures

Understanding the structural and functional characteristics of proteins are crucial for developing preventative and curative strategies that impact fields from drug discovery to policy development. An important and popular technique for examining how amino acids make up these characteristics of the protein sequences with position-specific scoring (PSS). While the string kernel is crucial in natural language processing (NLP), it is unclear if string kernels can extract biologically meaningful information from protein sequences, despite the fact that they have been shown to be effective in the general sequence analysis tasks. In this work, we propose a weighted PSS kernel matrix (or W-PSSKM), that combines a PSS representation of protein sequences, which encodes the frequency information of each amino acid in a sequence, with the notion of the string kernel. This results in a novel kernel function that outperforms many other approaches for protein sequence classification. We perform extensive experimentation to evaluate the proposed method. Our findings demonstrate that the W-PSSKM significantly outperforms existing baselines and state-of-the-art methods and achieves up to 45.1\% improvement in classification accuracy.

Dose selection is critical in pharmaceutical drug development, as it directly impacts therapeutic efficacy and patient safety of a drug. The Generalized Multiple Comparison Procedures and Modeling (MCP-Mod) approach is commonly used in Phase II trials for testing and estimation of dose-response relationships. However, its effectiveness in small sample sizes, particularly with binary endpoints, is hindered by issues like complete separation in logistic regression, leading to non-existence of estimates. Motivated by an actual clinical trial using the MCP-Mod approach, this paper introduces penalized maximum likelihood estimation (MLE) and randomization-based inference techniques to address these challenges. Randomization-based inference allows for exact finite sample inference, while population-based inference for MCP-Mod typically relies on asymptotic approximations. Simulation studies demonstrate that randomization-based tests can enhance statistical power in small to medium-sized samples while maintaining control over type-I error rates, even in the presence of time trends. Our results show that residual-based randomization tests using penalized MLEs not only improve computational efficiency but also outperform standard randomization-based methods, making them an adequate choice for dose-finding analyses within the MCP-Mod framework. Additionally, we apply these methods to pharmacometric settings, demonstrating their effectiveness in such scenarios. The results in this paper underscore the potential of randomization-based inference for the analysis of dose-finding trials, particularly in small sample contexts.

To study the dust dynamics in the dust trapping vortices in the protoplanetary disk around MWC~758, we analyzed the 1.3 mm continuum images of the MWC~758 disk obtained with the Atacama Large Millimeter/submillimeter Array (ALMA) in 2017 and 2021. We detect proper motions of 22 mas and 24 mas in the two dust clumps at radii of 0\farcs32 and 0\farcs54 in the disk on the plane of the sky, respectively. On the assumption that the dust clumps are located in the disk midplane, the velocities of the observed proper motions along the azimuthal direction of the inner and outer dust clumps are sub- and super-Keplerian, respectively, and both have angular velocities corresponding to the Keplerian angular velocity at a radius of $0\farcs46\pm0\farcs04$. This deviation from the Keplerian motion is not expected in the conventional theory of vortices formed by the Rossby wave instability. The observed non-Keplerian proper motions of the dust clumps are unlikely due to the disk warp and eccentricity, nor be associated with any predicted planets. The two dust clumps are likely spatially coincident with the infrared spirals. In addition, we detect the changes in the intensity profiles of the dust clumps over the four-year span. Therefore, we suggest that the observed proper motions are possibly due to changes in the density distributions in the dust clumps caused by their interaction with the spirals in the disk.

Training Large Language Models (LLMs) efficiently at scale presents a formidable challenge, driven by their ever-increasing computational demands and the need for enhanced performance. In this work, we introduce Liger-Kernel, an open-sourced set of Triton kernels developed specifically for LLM training. With kernel optimization techniques like kernel operation fusing and input chunking, our kernels achieve on average a 20% increase in training throughput and a 60% reduction in GPU memory usage for popular LLMs compared to HuggingFace implementations. In addition, Liger-Kernel is designed with modularity, accessibility, and adaptability in mind, catering to both casual and expert users. Comprehensive benchmarks and integration tests are built in to ensure compatibility, performance, correctness, and convergence across diverse computing environments and model architectures. The source code is available under a permissive license at: github.com/linkedin/Liger-Kernel., Comment: 17 pages, 12 figures

Stress tracing is an important research domain that supports many applications, such as health care and stress management; and its closest related works are derived from stress detection. However, these existing works cannot well address two important challenges facing stress detection. First, most of these studies involve asking users to wear physiological sensors to detect their stress states, which has a negative impact on the user experience. Second, these studies have failed to effectively utilize multimodal physiological signals, which results in less satisfactory detection results. This paper formally defines the stress tracing problem, which emphasizes the continuous detection of human stress states. A novel deep stress tracing method, named DST, is presented. Note that DST proposes tracing human stress based on physiological signals collected by a noncontact ultrawideband radar, which is more friendly to users when collecting their physiological signals. In DST, a signal extraction module is carefully designed at first to robustly extract multimodal physiological signals from the raw RF data of the radar, even in the presence of body movement. Afterward, a multimodal fusion module is proposed in DST to ensure that the extracted multimodal physiological signals can be effectively fused and utilized. Extensive experiments are conducted on three real-world datasets, including one self-collected dataset and two publicity datasets. Experimental results show that the proposed DST method significantly outperforms all the baselines in terms of tracing human stress states. On average, DST averagely provides a 6.31% increase in detection accuracy on all datasets, compared with the best baselines., Comment: 19 pages, 11 figures

Fine-tuning on task-specific data to boost downstream performance is a crucial step for leveraging Large Language Models (LLMs). However, previous studies have demonstrated that fine-tuning the models on several adversarial samples or even benign data can greatly comprise the model's pre-equipped alignment and safety capabilities. In this work, we propose SEAL, a novel framework to enhance safety in LLM fine-tuning. SEAL learns a data ranker based on the bilevel optimization to up rank the safe and high-quality fine-tuning data and down rank the unsafe or low-quality ones. Models trained with SEAL demonstrate superior quality over multiple baselines, with 8.5% and 9.7% win rate increase compared to random selection respectively on Llama-3-8b-Instruct and Merlinite-7b models. Our code is available on github https://github.com/hanshen95/SEAL.

Recent advances in Text-to-Speech (TTS) and Voice-Conversion (VC) using generative Artificial Intelligence (AI) technology have made it possible to generate high-quality and realistic human-like audio. This introduces significant challenges to distinguishing AI-synthesized speech from the authentic human voice and could raise potential issues of misuse for malicious purposes such as impersonation and fraud, spreading misinformation, deepfakes, and scams. However, existing detection techniques for AI-synthesized audio have not kept pace and often exhibit poor generalization across diverse datasets. In this paper, we introduce SONAR, a synthetic AI-Audio Detection Framework and Benchmark, aiming to provide a comprehensive evaluation for distinguishing cutting-edge AI-synthesized auditory content. SONAR includes a novel evaluation dataset sourced from 9 diverse audio synthesis platforms, including leading TTS providers and state-of-the-art TTS models. It is the first framework to uniformly benchmark AI-audio detection across both traditional and foundation model-based deepfake detection systems. Through extensive experiments, we reveal the generalization limitations of existing detection methods and demonstrate that foundation models exhibit stronger generalization capabilities, which can be attributed to their model size and the scale and quality of pretraining data. Additionally, we explore the effectiveness and efficiency of few-shot fine-tuning in improving generalization, highlighting its potential for tailored applications, such as personalized detection systems for specific entities or individuals. Code and dataset are available at https://github.com/Jessegator/SONAR.

Reducing the likelihood of generating harmful and toxic output is an essential task when aligning large language models (LLMs). Existing methods mainly rely on training an external reward model (i.e., another language model) or fine-tuning the LLM using self-generated data to influence the outcome. In this paper, we show that LLMs have the capability of self-detoxification without the use of an additional reward model or re-training. We propose \textit{Self-disciplined Autoregressive Sampling (SASA)}, a lightweight controlled decoding algorithm for toxicity reduction of LLMs. SASA leverages the contextual representations from an LLM to learn linear subspaces characterizing toxic v.s. non-toxic output in analytical forms. When auto-completing a response token-by-token, SASA dynamically tracks the margin of the current output to steer the generation away from the toxic subspace, by adjusting the autoregressive sampling strategy. Evaluated on LLMs of different scale and nature, namely Llama-3.1-Instruct (8B), Llama-2 (7B), and GPT2-L models with the RealToxicityPrompts, BOLD, and AttaQ benchmarks, SASA markedly enhances the quality of the generated sentences relative to the original models and attains comparable performance to state-of-the-art detoxification techniques, significantly reducing the toxicity level by only using the LLM's internal representations., Comment: 20 pages

Scaling laws in language modeling traditionally quantify training loss as a function of dataset size and model parameters, providing compute-optimal estimates but often neglecting the impact of data quality on model generalization. In this paper, we extend the conventional understanding of scaling law by offering a microscopic view of data quality within the original formulation -- effective training tokens -- which we posit to be a critical determinant of performance for parameter-constrained language models. Specifically, we formulate the proposed term of effective training tokens to be a combination of two readily-computed indicators of text: (i) text diversity and (ii) syntheticity as measured by a teacher model. We pretrained over $200$ models of 25M to 1.5B parameters on a diverse set of sampled, synthetic data, and estimated the constants that relate text quality, model size, training tokens, and eight reasoning task accuracy scores. We demonstrated the estimated constants yield +0.83 Pearson correlation with true accuracies, and analyzed it in scenarios involving widely-used data techniques such as data sampling and synthesis which aim to improve data quality., Comment: Accepted to EMNLP 2024 Industry Track, 18 pages, 9 figures, 4 tables

LLM-as-a-Judge has been widely utilized as an evaluation method in various benchmarks and served as supervised rewards in model training. However, despite their excellence in many domains, potential issues are under-explored, undermining their reliability and the scope of their utility. Therefore, we identify 12 key potential biases and propose a new automated bias quantification framework-CALM-which systematically quantifies and analyzes each type of bias in LLM-as-a-Judge by using automated and principle-guided modification. Our experiments cover multiple popular language models, and the results indicate that while advanced models have achieved commendable overall performance, significant biases persist in certain specific tasks. Empirical results suggest that there remains room for improvement in the reliability of LLM-as-a-Judge. Moreover, we also discuss the explicit and implicit influence of these biases and give some suggestions for the reliable application of LLM-as-a-Judge. Our work highlights the need for stakeholders to address these issues and remind users to exercise caution in LLM-as-a-Judge applications.

Chain-of-Thought (CoT) is an efficient prompting method that enables the reasoning ability of large language models by augmenting the query using multiple examples with multiple intermediate steps. Despite the empirical success, the theoretical understanding of how to train a Transformer to achieve the CoT ability remains less explored. This is primarily due to the technical challenges involved in analyzing the nonconvex optimization on nonlinear attention models. To the best of our knowledge, this work provides the first theoretical study of training Transformers with nonlinear attention to obtain the CoT generalization capability so that the resulting model can inference on unseen tasks when the input is augmented by examples of the new task. We first quantify the required training samples and iterations to train a Transformer model towards CoT ability. We then prove the success of its CoT generalization on unseen tasks with distribution-shifted testing data. Moreover, we theoretically characterize the conditions for an accurate reasoning output by CoT even when the provided reasoning examples contain noises and are not always accurate. In contrast, in-context learning (ICL), which can be viewed as one-step CoT without intermediate steps, may fail to provide an accurate output when CoT does. These theoretical findings are justified through experiments.

This paper systematically investigates the quasinormal modes (QNMs) of a scalar field around a charged loop quantum black hole (BH). The BH is characterized by two key parameters: the quantum parameter $b_0$ and the charge parameter $Q$. We explore two scenarios: one where the multipole quantum number $l=0$, in which only the gravitational potential is relevant, and another where $l > 0$, in which both $Q$ and $b_0$ jointly influence the quasinormal frequencies (QNFs). For $l=0$, quantum gravity effects lead to a pronounced overtone outburst in the QNFs, with the outburst and accompanying oscillatory patterns becoming more pronounced as the overtone number increases. For $l > 0$, the overtone outburst disappears due to the centrifugal potential's suppressive the quantum gravity effect. However, a non-monotonic behavior in the first two overtones with respect to $b_0$ is observed, indicating that subtle quantum gravity effects persist. This behavior vanishes as $Q$ increases, further confirming the suppressive influence of the centrifugal potential on the quantum gravity effect.

Cardiovascular disease (CVD) is the leading cause of death and premature mortality worldwide, with occupational environments significantly influencing CVD risk, underscoring the need for effective cardiac monitoring and early warning systems. Existing methods of monitoring vital signs require subjects to remain stationary, which is impractical for daily monitoring as individuals are often in motion. To address this limitation, we propose MERIT, a multimodality-based wearable system designed for precise ECG waveform monitoring without movement restrictions. Daily activities, involving frequent arm movements, can significantly affect sensor data and complicate the reconstruction of accurate ECG signals. To mitigate motion impact and enhance ECG signal reconstruction, we introduce a deep independent component analysis (Deep-ICA) module and a multimodal fusion module. We conducted experiments with 15 subjects. Our results, compared with commercial wearable devices and existing methods, demonstrate that MERIT accurately reconstructs ECG waveforms during various office activities, offering a reliable solution for fine-grained cardiac monitoring in dynamic environments., Comment: 8 pages, 10 figures

In this work, we propose a novel consistency-preserving loss function for recovering the phase information in the context of phase reconstruction (PR) and speech enhancement (SE). Different from conventional techniques that directly estimate the phase using a deep model, our idea is to exploit ad-hoc constraints to directly generate a consistent pair of magnitude and phase. Specifically, the proposed loss forces a set of complex numbers to be a consistent short-time Fourier transform (STFT) representation, i.e., to be the spectrogram of a real signal. Our approach thus avoids the difficulty of estimating the original phase, which is highly unstructured and sensitive to time shift. The influence of our proposed loss is first assessed on a PR task, experimentally demonstrating that our approach is viable. Next, we show its effectiveness on an SE task, using both the VB-DMD and WSJ0-CHiME3 data sets. On VB-DMD, our approach is competitive with conventional solutions. On the challenging WSJ0-CHiME3 set, the proposed framework compares favourably over those techniques that explicitly estimate the phase., Comment: 5 pages, Submitted to ICASSP 2025

This paper proposes a generative pretraining foundation model for high-quality speech restoration tasks. By directly operating on complex-valued short-time Fourier transform coefficients, our model does not rely on any vocoders for time-domain signal reconstruction. As a result, our model simplifies the synthesis process and removes the quality upper-bound introduced by any mel-spectrogram vocoder compared to prior work SpeechFlow. The proposed method is evaluated on multiple speech restoration tasks, including speech denoising, bandwidth extension, codec artifact removal, and target speaker extraction. In all scenarios, finetuning our pretrained model results in superior performance over strong baselines. Notably, in the target speaker extraction task, our model outperforms existing systems, including those leveraging SSL-pretrained encoders like WavLM. The code and the pretrained checkpoints are publicly available in the NVIDIA NeMo framework., Comment: 5 pages, Submitted to ICASSP 2025. The implementation and configuration could be found in https://github.com/NVIDIA/NeMo/blob/main/examples/audio/conf/flow_matching_generative_ssl_pretraining.yaml The audio demo page could be found in https://kuray107.github.io/ssl_gen25-examples/index.html

As generative AI, particularly large language models (LLMs), become increasingly integrated into production applications, new attack surfaces and vulnerabilities emerge and put a focus on adversarial threats in natural language and multi-modal systems. Red-teaming has gained importance in proactively identifying weaknesses in these systems, while blue-teaming works to protect against such adversarial attacks. Despite growing academic interest in adversarial risks for generative AI, there is limited guidance tailored for practitioners to assess and mitigate these challenges in real-world environments. To address this, our contributions include: (1) a practical examination of red- and blue-teaming strategies for securing generative AI, (2) identification of key challenges and open questions in defense development and evaluation, and (3) the Attack Atlas, an intuitive framework that brings a practical approach to analyzing single-turn input attacks, placing it at the forefront for practitioners. This work aims to bridge the gap between academic insights and practical security measures for the protection of generative AI systems.

Language model pretraining generally targets a broad range of use cases and incorporates data from diverse sources. However, there are instances where we desire a model that excels in specific areas without markedly compromising performance in other areas. A cost-effective and straightforward approach is sampling with low-dimensional data features, which allows to select large-scale pretraining data for domain-specific use cases. In this work, we revisit importance sampling with n-gram features consisting of multi-granular tokens, which strikes a good balance between sentence compression and representation capabilities. We observed the sampled data to have a high correlation with the target downstream task performance while preserving its effectiveness on other tasks. This leads to the proposed data sampling paradigm where language models can be pretrained more efficiently on selected documents. On eight benchmarks we demonstrate with $\sim$1% of the data, pretrained models perform on par with the full RefinedWeb data and outperform randomly selected samples for model sizes ranging from 125M to 1.5B., Comment: Accepted to EMNLP 2024 Main Conference, 9 pages, 6 figures, 3 tables

We present experiments to reproduce the characteristics of core-collapse supernovae with different stellar masses and initial explosion energies in the laboratory. In the experiments, shocks are driven in 1.2 atm and 1.9 atm xenon gas by laser with energy from 1600J to 2800J on the SGIII prototype laser facility. The average shock velocities and shocked densities are obtained from experiments. Experimental results reveal that higher laser energy and lower Xe gas density led to higher shock velocity, and lower Xe gas initial density has a higher compression. Modeling of the experiments using the 2D radiation hydrodynamic codes Icefire shows excellent agreement with the experimental results and gives the temperature. These results will contribute to time-domain astrophysical systems, such as gravitational supernovae, where a strong radiative shock propagates outward from the center of the star after the core collapses., Comment: 7 pages, 2 figures, 1 supplement (8 pages, 3 figures, 2 tables), accepted for publication in Science Bulletin

Recent advances in VLSI fabrication technology have led to die shrinkage and increased layout density, creating an urgent demand for advanced hotspot detection techniques. However, by taking an object detection network as the backbone, recent learning-based hotspot detectors learn to recognize only the problematic layout patterns in the training data. This fact makes these hotspot detectors difficult to generalize to real-world scenarios. We propose a novel lithography simulator-powered hotspot detection framework to overcome this difficulty. Our framework integrates a lithography simulator with an object detection backbone, merging the extracted latent features from both the simulator and the object detector via well-designed cross-attention blocks. Consequently, the proposed framework can be used to detect potential hotspot regions based on I) the variation of possible circuit shape deformation estimated by the lithography simulator, and ii) the problematic layout patterns already known. To this end, we utilize RetinaNet with a feature pyramid network as the object detection backbone and leverage LithoNet as the lithography simulator. Extensive experiments demonstrate that our proposed simulator-guided hotspot detection framework outperforms previous state-of-the-art methods on real-world data., Comment: 14 pages to appear in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems

In this paper, we investigate the impact of loop quantum gravity (LQG) on extreme mass-ratio inspirals (EMRIs), and the results indicate that LQG effects cause the orbital decay to occur faster compared to the Schwarzschild case. Furthermore, we use the augmented analytic kludge approach to generate EMRI waveforms and study the LISA's capability to detect the LQG effect with faithfulness. Additionally, employing the Fisher information matrix method for parameter estimation, we estimate that after one year of observation, the uncertainty in $r_0$ reduces to approximately $6.59\times 10^{-4}$ with a signal-to-noise ratio of $49$., Comment: 22 pages,7 figures

One of the primary science goals of the Habitable Worlds Observatory (HWO) as defined by the Astro2020 decadal survey is the imaging of the first Earth-like planet around a Sun-like star. A key technology gap towards reaching this goal are the development of ultra-low-noise photon counting detectors capable of measuring the incredibly low count rates coming from these planets which are at contrasts of $\sim 1 \times 10^{-10}$. Superconducting energy-resolving detectors (ERDs) are a promising technology for this purpose as, despite their technological challenges, needing to be cooled below their superconducting transition temperature ($< 1\mathrm{K}$), they have essentially zero read noise, dark current, or clock-induced charge, and can get the wavelength of each incident photon without the use of additional throughput-reducing filters or gratings that spread light over many pixels. The use of these detectors on HWO will not only impact the science of the mission by decreasing the required exposure times for exo-Earth detection and characterization, but also in a wavefront sensing and control context when used for starlight suppression to generate a dark zone. We show simulated results using both an EMCCD and an ERD to ``dig a dark zone'' demonstrating that ERDs can achieve the same final contrast as an EMCCD in about half of the total time. We also perform a simple case study using an exposure time calculator tool called the Error Budget Software (EBS) to determine the required integration times to detect water for HWO targets of interest using both EMCCDs and ERDs. This shows that once a dark zone is achieved, using an ERD can decrease these exposure times by factors of 1.5--2 depending on the specific host star properties., Comment: 13 pages, 7 figures

LiDAR-based 3D object detection is a critical technology for the development of autonomous driving and robotics. However, the high cost of data annotation limits its advancement. We propose a novel and effective active learning (AL) method called Distribution Discrepancy and Feature Heterogeneity (DDFH), which simultaneously considers geometric features and model embeddings, assessing information from both the instance-level and frame-level perspectives. Distribution Discrepancy evaluates the difference and novelty of instances within the unlabeled and labeled distributions, enabling the model to learn efficiently with limited data. Feature Heterogeneity ensures the heterogeneity of intra-frame instance features, maintaining feature diversity while avoiding redundant or similar instances, thus minimizing annotation costs. Finally, multiple indicators are efficiently aggregated using Quantile Transform, providing a unified measure of informativeness. Extensive experiments demonstrate that DDFH outperforms the current state-of-the-art (SOTA) methods on the KITTI and Waymo datasets, effectively reducing the bounding box annotation cost by 56.3% and showing robustness when working with both one-stage and two-stage models., Comment: Accepted to CoRL 2024

Due to the diverse geographical environments, intricate landscapes, and high-density settlements, the automatic identification of urban village boundaries using remote sensing images remains a highly challenging task. This paper proposes a novel and efficient neural network model called UV-Mamba for accurate boundary detection in high-resolution remote sensing images. UV-Mamba mitigates the memory loss problem in lengthy sequence modeling, which arises in state space models with increasing image size, by incorporating deformable convolutions. Its architecture utilizes an encoder-decoder framework and includes an encoder with four deformable state space augmentation blocks for efficient multi-level semantic extraction and a decoder to integrate the extracted semantic information. We conducted experiments on two large datasets showing that UV-Mamba achieves state-of-the-art performance. Specifically, our model achieves 73.3% and 78.1% IoU on the Beijing and Xi'an datasets, respectively, representing improvements of 1.2% and 3.4% IoU over the previous best model while also being 6x faster in inference speed and 40x smaller in parameter count. Source code and pre-trained models are available at https://github.com/Devin-Egber/UV-Mamba., Comment: 5 pages, 4 figures, 3 tables

Adapting pre-trained models to new tasks can exhibit varying effectiveness across datasets. Visual prompting, a state-of-the-art parameter-efficient transfer learning method, can significantly improve the performance of out-of-distribution tasks. On the other hand, linear probing, a standard transfer learning method, can sometimes become the best approach. We propose a log-likelihood ratio (LLR) approach to analyze the comparative benefits of visual prompting and linear probing. By employing the LLR score alongside resource-efficient visual prompts approximations, our cost-effective measure attains up to a 100-fold reduction in run time compared to full training, while achieving prediction accuracies up to 91%. The source code is available at https://github.com/IBM/VP-LLR.

Electrical manipulation of spin-polarized current is highly desirable yet tremendously challenging in developing ultracompact spintronic device technology. Here we propose a scheme to realize the all-electrical manipulation of spin-polarized current in an altermagnetic bilayer. Such a bilayer system can host layer-spin locking, in which one layer hosts a spin-polarized current while the other layer hosts a current with opposite spin polarization. An out-of-plane electric field breaks the layer degeneracy, leading to a gate-tunable spin-polarized current whose polarization can be fully reversed upon flipping the polarity of the electric field. Using first-principles calculations, we show that CrS bilayer with C-type antiferromagnetic exchange interaction exhibits a hidden layer-spin locking mechanism that enables the spin polarization of the transport current to be electrically manipulated via the layer degree of freedom. We demonstrate that sign-reversible spin polarization as high as 87% can be achieved at room temperature. This work presents the pioneering concept of layer-spintronics which synergizes altermagnetism and bilayer stacking to achieve efficient electrical control of spin., Comment: 20 pages, 5 figures

Established sampling protocols for 3D point cloud learning, such as Farthest Point Sampling (FPS) and Fixed Sample Size (FSS), have long been recognized and utilized. However, real-world data often suffer from corrputions such as sensor noise, which violates the benignness assumption of point cloud in current protocols. Consequently, they are notably vulnerable to noise, posing significant safety risks in critical applications like autonomous driving. To address these issues, we propose an enhanced point cloud sampling protocol, PointDR, which comprises two components: 1) Downsampling for key point identification and 2) Resampling for flexible sample size. Furthermore, differentiated strategies are implemented for training and inference processes. Particularly, an isolation-rated weight considering local density is designed for the downsampling method, assisting it in performing random key points selection in the training phase and bypassing noise in the inference phase. A local-geometry-preserved upsampling is incorporated into resampling, facilitating it to maintain a stochastic sample size in the training stage and complete insufficient data in the inference. It is crucial to note that the proposed protocol is free of model architecture altering and extra learning, thus minimal efforts are demanded for its replacement of the existing one. Despite the simplicity, it substantially improves the robustness of point cloud learning, showcased by outperforming the state-of-the-art methods on multiple benchmarks of corrupted point cloud classification. The code will be available upon the paper's acceptance.

Most evaluations of vision models use indirect methods to assess latent space quality. These methods often involve adding extra layers to project the latent space into a new one. This projection makes it difficult to analyze and compare the original latent space. This article uses the k* Distribution, a local neighborhood analysis method, to examine the learned latent space at the level of individual concepts, which can be extended to examine the entire latent space. We introduce skewness-based true and approximate metrics for interpreting individual concepts to assess the overall quality of vision models' latent space. Our findings indicate that current vision models frequently fracture the distributions of individual concepts within the latent space. Nevertheless, as these models improve in generalization across multiple datasets, the degree of fracturing diminishes. A similar trend is observed in robust vision models, where increased robustness correlates with reduced fracturing. Ultimately, this approach enables a direct interpretation and comparison of the latent spaces of different vision models and reveals a relationship between a model's generalizability and robustness. Results show that as a model becomes more general and robust, it tends to learn features that result in better clustering of concepts. Project Website is available online at https://shashankkotyan.github.io/k-Distribution/

We present here the current state of a collection of promising ultraviolet technologies in preparation for the Habitable Worlds Observatory. Working with experts representing a significant number of groups working in the ultraviolet, we summarize some of the leading science drivers, present an argument for a 100 nm blue wavelength cutoff, and gather current state of the art of UV technologies. We present the state of the art of contamination control, a crucial piece of the UV instrument plan. We explore next steps with individual technologies, as well as present paths forward with systems level testing and development.

Piano cover generation aims to create a piano cover from a pop song. Existing approaches mainly employ supervised learning and the training demands strongly-aligned and paired song-to-piano data, which is built by remapping piano notes to song audio. This would, however, result in the loss of piano information and accordingly cause inconsistencies between the original and remapped piano versions. To overcome this limitation, we propose a transfer learning approach that pre-trains our model on piano-only data and fine-tunes it on weakly-aligned paired data constructed without note remapping. During pre-training, to guide the model to learn piano composition concepts instead of merely transcribing audio, we use an existing lead sheet transcription model as the encoder to extract high-level features from the piano recordings. The pre-trained model is then fine-tuned on the paired song-piano data to transfer the learned composition knowledge to the pop song domain. Our evaluation shows that this training strategy enables our model, named PiCoGen2, to attain high-quality results, outperforming baselines on both objective and subjective metrics across five pop genres., Comment: Accepted at the 25th International Society for Music Information Retrieval Conference (ISMIR), 2024

Cover song generation stands out as a popular way of music making in the music-creative community. In this study, we introduce Piano Cover Generation (PiCoGen), a two-stage approach for automatic cover song generation that transcribes the melody line and chord progression of a song given its audio recording, and then uses the resulting lead sheet as the condition to generate a piano cover in the symbolic domain. This approach is advantageous in that it does not required paired data of covers and their original songs for training. Compared to an existing approach that demands such paired data, our evaluation shows that PiCoGen demonstrates competitive or even superior performance across songs of different musical genres., Comment: Published at ICMR 2024 (project page: https://tanchihpin0517.github.io/PiCoGen/)

As 5G networks strive to support advanced time-critical applications, such as immersive Extended Reality (XR), cloud gaming, and autonomous driving, the demand for Real-time Broadband Communication (RTBC) grows. In this article, we present the main mechanisms of Low Latency, Low Loss, and Scalable Throughput (L4S). Subsequently, we investigate the support and challenges of L4S technology in the latest 3GPP 5G-Advanced Release 18 (R18) standard. Our case study, using a prototype system for a real-time communication (RTC) application, demonstrates the superiority of L4S technology. The experimental results show that, compared with the GCC algorithm, the proposed L4S-GCC algorithm can reduce the stalling rate by 1.51%-2.80% and increase the bandwidth utilization by 11.4%-31.4%. The results emphasize the immense potential of L4S technology in enhancing transmission performance in time-critical applications., Comment: 7 pages, 3 figures. This work has been submitted to the IEEE for possible publication

Sepsis is the leading cause of in-hospital mortality in the USA. Early sepsis onset prediction and diagnosis could significantly improve the survival of sepsis patients. Existing predictive models are usually trained on high-quality data with few missing information, while missing values widely exist in real-world clinical scenarios (especially in the first hours of admissions to the hospital), which causes a significant decrease in accuracy and an increase in uncertainty for the predictive models. The common method to handle missing values is imputation, which replaces the unavailable variables with estimates from the observed data. The uncertainty of imputation results can be propagated to the sepsis prediction outputs, which have not been studied in existing works on either sepsis prediction or uncertainty quantification. In this study, we first define such propagated uncertainty as the variance of prediction output and then introduce uncertainty propagation methods to quantify the propagated uncertainty. Moreover, for the potential high-risk patients with low confidence due to limited observations, we propose a robust active sensing algorithm to increase confidence by actively recommending clinicians to observe the most informative variables. We validate the proposed models in both publicly available data (i.e., MIMIC-III and AmsterdamUMCdb) and proprietary data in The Ohio State University Wexner Medical Center (OSUWMC). The experimental results show that the propagated uncertainty is dominant at the beginning of admissions to hospitals and the proposed algorithm outperforms state-of-the-art active sensing methods. Finally, we implement a SepsisLab system for early sepsis prediction and active sensing based on our pre-trained models. Clinicians and potential sepsis patients can benefit from the system in early prediction and diagnosis of sepsis., Comment: To be published in KDD 2024

Identifying beneficial tasks to transfer from is a critical step toward successful intermediate-task transfer learning. In this work, we experiment with 130 source-target task combinations and demonstrate that the transfer performance exhibits severe variance across different source tasks and training seeds, highlighting the crucial role of intermediate-task selection in a broader context. We compare four representative task selection methods in a unified setup, focusing on their effectiveness and consistency. Compared to embedding-free methods and text embeddings, task embeddings constructed from fine-tuned weights can better estimate task transferability by improving task prediction scores from 2.59% to 3.96%. Despite their strong performance, we observe that the task embeddings do not consistently demonstrate superiority for tasks requiring reasoning abilities. Furthermore, we introduce a novel method that measures pairwise token similarity using maximum inner product search, leading to the highest performance in task prediction. Our findings suggest that token-wise similarity is better predictive for predicting transferability compared to averaging weights., Comment: Accepted to ACL SRW 2024

With the development of artificial intelligence, human beings are increasingly interested in human-agent collaboration, which generates a series of problems about the relationship between agents and humans, such as trust and cooperation. This inevitably induces the inherent human characteristic that there are subjective interaction preferences for different groups, especially in human-agent hybrid systems where human-human interaction, agent-agent interaction, and human-agent interaction coexist. However, understanding how individual interaction preferences affect the cooperation of the system remains a major challenge. Therefore, this paper proposes a human-agent hybrid prisoner's dilemma game system under the framework of evolutionary game. In spatial networks, the most significant difference between agents and humans is the flexibility of decision, where humans have higher adaptive capabilities, follow link dynamics, and adopt free decision rules, which enable them to choose different strategies for different neighbors. However, agents follow node dynamics and adopt consistent decision rules, applying the same strategy to different neighbors. We give the subjective preferences of any individual to different groups, involving the interaction preferences between homogeneous groups and heterogeneous groups respectively. The simulation results show that both human and agent have asymmetric interaction preferences for groups with different identities, which can significantly improve the cooperative behavior of the system. In the hybrid system, human groups show more stable prosocial behavior. Agent groups can form highly cooperative clusters under the condition of strong interaction preference for human groups. In addition, giving agents the ability to identify opponents can effectively alleviate the interaction dilemma of agents.

Perceiving the world as 3D occupancy supports embodied agents to avoid collision with any types of obstacle. While open-vocabulary image understanding has prospered recently, how to bind the predicted 3D occupancy grids with open-world semantics still remains under-explored due to limited open-world annotations. Hence, instead of building our model from scratch, we try to blend 2D foundation models, specifically a depth model MiDaS and a semantic model CLIP, to lift the semantics to 3D space, thus fulfilling 3D occupancy. However, building upon these foundation models is not trivial. First, the MiDaS faces the depth ambiguity problem, i.e., it only produces relative depth but fails to estimate bin depth for feature lifting. Second, the CLIP image features lack high-resolution pixel-level information, which limits the 3D occupancy accuracy. Third, open vocabulary is often trapped by the long-tail problem. To address these issues, we propose VEON for Vocabulary-Enhanced Occupancy predictioN by not only assembling but also adapting these foundation models. We first equip MiDaS with a Zoedepth head and low-rank adaptation (LoRA) for relative-metric-bin depth transformation while reserving beneficial depth prior. Then, a lightweight side adaptor network is attached to the CLIP vision encoder to generate high-resolution features for fine-grained 3D occupancy prediction. Moreover, we design a class reweighting strategy to give priority to the tail classes. With only 46M trainable parameters and zero manual semantic labels, VEON achieves 15.14 mIoU on Occ3D-nuScenes, and shows the capability of recognizing objects with open-vocabulary categories, meaning that our VEON is label-efficient, parameter-efficient, and precise enough., Comment: Accepted to ECCV2024

Cavity expansion is a canonical problem in geotechnics, which can be described by partial differential equations (PDEs) and ordinary differential equations (ODEs). This study explores the potential of using a new solver, a physics-informed neural network (PINN), to calculate the stress field in an expanded cavity in the elastic and elasto-plastic regimes. Whilst PINNs have emerged as an effective universal function approximator for deriving the solutions of a wide range of governing PDEs/ODEs, their ability to solve elasto-plastic problems remains uncertain. A novel parsimonious loss function is first proposed to balance the simplicity and accuracy of PINN. The proposed method is applied to diverse material behaviours in the cavity expansion problem including isotropic, anisotropic elastic media, and elastic-perfectly plastic media with Tresca and Mohr-Coulomb yield criteria. The results indicate that the use of a parsimonious prior information-based loss function is highly beneficial to deriving the approximate solutions of complex PDEs with high accuracy. The present method allows for accurate derivation of solutions for both elastic and plastic mechanical responses of an expanded cavity. It also provides insights into how PINNs can be further advanced to solve more complex problems in geotechnical practice.

We analytically compute the power spectrum of primordial curvature perturbations in Type III hilltop inflation models under the slow-roll approximation. The model parameters are constrained using current Cosmic Microwave Background (CMB) data. The curvature perturbations that exit the horizon at small scales show sufficiently large amplitudes to produce primordial black holes (PBHs). We then consider the quantum one-loop corrections in these models from both the self-interaction of the inflaton and its interaction with the waterfall field. We show the loop corrections in both cases for 60 e-folds of inflation are negligible, ensuring the tree-level results are reliable within the chosen parameter regime., Comment: 13 pages, 6 figures. arXiv admin note: text overlap with arXiv:2311.13746

Multi-modal large language models (MLLMs) have been given free rein to explore exciting medical applications with a primary focus on radiology report generation. Nevertheless, the preliminary success in 2D radiology captioning is incompetent to reflect the real-world diagnostic challenge in the volumetric 3D anatomy. To mitigate three crucial limitation aspects in the existing literature, including (1) data complexity, (2) model capacity, and (3) evaluation metric fidelity, we collected an 18,885 text-scan pairs 3D-BrainCT dataset and applied clinical visual instruction tuning (CVIT) to train BrainGPT models to generate radiology-adherent 3D brain CT reports. Statistically, our BrainGPT scored BLEU-1 = 44.35, BLEU-4 = 20.38, METEOR = 30.13, ROUGE-L = 47.6, and CIDEr-R = 211.77 during internal testing and demonstrated an accuracy of 0.91 in captioning midline shifts on the external validation CQ500 dataset. By further inspecting the captioned report, we reported that the traditional metrics appeared to measure only the surface text similarity and failed to gauge the information density of the diagnostic purpose. To close this gap, we proposed a novel Feature-Oriented Radiology Task Evaluation (FORTE) to estimate the report's clinical relevance (lesion feature and landmarks). Notably, the BrainGPT model scored an average FORTE F1-score of 0.71 (degree=0.661; landmark=0.706; feature=0.693; impression=0.779). To demonstrate that BrainGPT models possess objective readiness to generate human-like radiology reports, we conducted a Turing test that enrolled 11 physician evaluators, and around 74% of the BrainGPT-generated captions were indistinguishable from those written by humans. Our work embodies a holistic framework that showcased the first-hand experience of curating a 3D brain CT dataset, fine-tuning anatomy-sensible language models, and proposing robust radiology evaluation metrics., Comment: 6 figures, 5 supplementary figures, 8 supplementary tables

Differential privacy (DP) is a widely used approach for mitigating privacy risks when training machine learning models on sensitive data. DP mechanisms add noise during training to limit the risk of information leakage. The scale of the added noise is critical, as it determines the trade-off between privacy and utility. The standard practice is to select the noise scale in terms of a privacy budget parameter $\epsilon$. This parameter is in turn interpreted in terms of operational attack risk, such as accuracy, or sensitivity and specificity of inference attacks against the privacy of the data. We demonstrate that this two-step procedure of first calibrating the noise scale to a privacy budget $\epsilon$, and then translating $\epsilon$ to attack risk leads to overly conservative risk assessments and unnecessarily low utility. We propose methods to directly calibrate the noise scale to a desired attack risk level, bypassing the intermediate step of choosing $\epsilon$. For a target attack risk, our approach significantly decreases noise scale, leading to increased utility at the same level of privacy. We empirically demonstrate that calibrating noise to attack sensitivity/specificity, rather than $\epsilon$, when training privacy-preserving ML models substantially improves model accuracy for the same risk level. Our work provides a principled and practical way to improve the utility of privacy-preserving ML without compromising on privacy.

The security and robustness of deep neural networks (DNNs) have become increasingly concerning. This paper aims to provide both a theoretical foundation and a practical solution to ensure the reliability of DNNs. We explore the concept of Lipschitz continuity to certify the robustness of DNNs against adversarial attacks, which aim to mislead the network with adding imperceptible perturbations into inputs. We propose a novel algorithm that remaps the input domain into a constrained range, reducing the Lipschitz constant and potentially enhancing robustness. Unlike existing adversarially trained models, where robustness is enhanced by introducing additional examples from other datasets or generative models, our method is almost cost-free as it can be integrated with existing models without requiring re-training. Experimental results demonstrate the generalizability of our method, as it can be combined with various models and achieve enhancements in robustness. Furthermore, our method achieves the best robust accuracy for CIFAR10, CIFAR100, and ImageNet datasets on the RobustBench leaderboard.

Efficient training and inference algorithms, such as low-rank adaption and model pruning, have shown impressive performance for learning Transformer-based large foundation models. However, due to the technical challenges of the non-convex optimization caused by the complicated architecture of Transformers, the theoretical study of why these methods can be applied to learn Transformers is mostly elusive. To the best of our knowledge, this paper shows the first theoretical analysis of the property of low-rank and sparsity of one-layer Transformers by characterizing the trained model after convergence using stochastic gradient descent. By focusing on a data model based on label-relevant and label-irrelevant patterns, we quantify that the gradient updates of trainable parameters are low-rank, which depends on the number of label-relevant patterns. We also analyze how model pruning affects the generalization while improving computation efficiency and conclude that proper magnitude-based pruning has a slight effect on the testing performance. We implement numerical experiments to support our findings., Comment: IEEE SAM Workshop 2024

Pre-trained Language models (PLMs) have been acknowledged to contain harmful information, such as social biases, which may cause negative social impacts or even bring catastrophic results in application. Previous works on this problem mainly focused on using black-box methods such as probing to detect and quantify social biases in PLMs by observing model outputs. As a result, previous debiasing methods mainly finetune or even pre-train language models on newly constructed anti-stereotypical datasets, which are high-cost. In this work, we try to unveil the mystery of social bias inside language models by introducing the concept of {\sc Social Bias Neurons}. Specifically, we propose {\sc Integrated Gap Gradients (IG$^2$)} to accurately pinpoint units (i.e., neurons) in a language model that can be attributed to undesirable behavior, such as social bias. By formalizing undesirable behavior as a distributional property of language, we employ sentiment-bearing prompts to elicit classes of sensitive words (demographics) correlated with such sentiments. Our IG$^2$ thus attributes the uneven distribution for different demographics to specific Social Bias Neurons, which track the trail of unwanted behavior inside PLM units to achieve interoperability. Moreover, derived from our interpretable technique, {\sc Bias Neuron Suppression (BNS)} is further proposed to mitigate social biases. By studying BERT, RoBERTa, and their attributable differences from debiased FairBERTa, IG$^2$ allows us to locate and suppress identified neurons, and further mitigate undesired behaviors. As measured by prior metrics from StereoSet, our model achieves a higher degree of fairness while maintaining language modeling ability with low cost.

Despite notable advancements in the field of computer vision, the precise detection of tiny objects continues to pose a significant challenge, largely owing to the minuscule pixel representation allocated to these objects in imagery data. This challenge resonates profoundly in the domain of geoscience and remote sensing, where high-fidelity detection of tiny objects can facilitate a myriad of applications ranging from urban planning to environmental monitoring. In this paper, we propose a new framework, namely, DeNoising FPN with Trans R-CNN (DNTR), to improve the performance of tiny object detection. DNTR consists of an easy plug-in design, DeNoising FPN (DN-FPN), and an effective Transformer-based detector, Trans R-CNN. Specifically, feature fusion in the feature pyramid network is important for detecting multiscale objects. However, noisy features may be produced during the fusion process since there is no regularization between the features of different scales. Therefore, we introduce a DN-FPN module that utilizes contrastive learning to suppress noise in each level's features in the top-down path of FPN. Second, based on the two-stage framework, we replace the obsolete R-CNN detector with a novel Trans R-CNN detector to focus on the representation of tiny objects with self-attention. Experimental results manifest that our DNTR outperforms the baselines by at least 17.4% in terms of APvt on the AI-TOD dataset and 9.6% in terms of AP on the VisDrone dataset, respectively. Our code will be available at https://github.com/hoiliu-0801/DNTR., Comment: The article is accepted by IEEE Transactions on Geoscience and Remote Sensing. Our code will be available at https://github.com/hoiliu-0801/DNTR

Deep neural networks are susceptible to backdoor attacks, where adversaries manipulate model predictions by inserting malicious samples into the training data. Currently, there is still a lack of direct filtering methods for identifying suspicious training data to unveil potential backdoor samples. In this paper, we propose a novel method, Prediction Shift Backdoor Detection (PSBD), leveraging an uncertainty-based approach requiring minimal unlabeled clean validation data. PSBD is motivated by an intriguing Prediction Shift (PS) phenomenon, where poisoned models' predictions on clean data often shift away from true labels towards certain other labels with dropout applied during inference, while backdoor samples exhibit less PS. We hypothesize PS results from neuron bias effect, making neurons favor features of certain classes. PSBD identifies backdoor training samples by computing the Prediction Shift Uncertainty (PSU), the variance in probability values when dropout layers are toggled on and off during model inference. Extensive experiments have been conducted to verify the effectiveness and efficiency of PSBD, which achieves state-of-the-art results among mainstream detection methods.