Search

The self-attention mechanism traditionally relies on the softmax operator, necessitating positional embeddings like RoPE, or position biases to account for token order. But current methods using still face length generalisation challenges. We propose an alternative attention mechanism based on the stick-breaking process: For each token before the current, we determine a break point $\beta_{i,j}$, which represents the proportion of the remaining stick to allocate to the current token. We repeat the process until the stick is fully allocated, resulting in a sequence of attention weights. This process naturally incorporates recency bias, which has linguistic motivations for grammar parsing (Shen et. al., 2017). We study the implications of replacing the conventional softmax-based attention mechanism with stick-breaking attention. We then discuss implementation of numerically stable stick-breaking attention and adapt Flash Attention to accommodate this mechanism. When used as a drop-in replacement for current softmax+RoPE attention systems, we find that stick-breaking attention performs competitively with current methods on length generalisation and downstream tasks. Stick-breaking also performs well at length generalisation, allowing a model trained with $2^{11}$ context window to perform well at $2^{14}$ with perplexity improvements.

Full-duplex spoken dialogue systems significantly advance over traditional turn-based dialogue systems, as they allow simultaneous bidirectional communication, closely mirroring human-human interactions. However, achieving low latency and natural interactions in full-duplex dialogue systems remains a significant challenge, especially considering human conversation dynamics such as interruptions, backchannels, and overlapping speech. In this paper, we introduce a novel End-to-End GPT-based model OmniFlatten for full-duplex conversation, capable of effectively modeling the complex behaviors inherent to natural conversations with low latency. To achieve full-duplex communication capabilities, we propose a multi-stage post-training scheme that progressively adapts a text-based large language model (LLM) backbone into a speech-text dialogue LLM, capable of generating text and speech in real time, without modifying the architecture of the backbone LLM. The training process comprises three stages: modality alignment, half-duplex dialogue learning, and full-duplex dialogue learning. Throughout all training stages, we standardize the data using a flattening operation, which allows us to unify the training methods and the model architecture across different modalities and tasks. Our approach offers a straightforward modeling technique and a promising research direction for developing efficient and natural end-to-end full-duplex spoken dialogue systems. Audio samples of dialogues generated by OmniFlatten can be found at this web site (https://omniflatten.github.io/)., Comment: Work in progress

Forming single-photon emitters (SPEs) in insulating hexagonal boron nitride (hBN) has sparked wide interests in the quantum photonics. Despite significant progress, it remains challenging to deterministically create SPEs at precise locations with a specific type of element for creating defects. In this study, we present a straightforward approach to generate site-deterministic carbon-functionalized quantum emitters in hBN by harnessing ultrasonic nanoindentation. The obtained SPEs are high-quality and can be scaled up to large arrays in a single fabrication step. Comprehensive experimental analyses reveal that the insertion of carbon atoms into the hBN lattice is the source of the robust quantum emission. Complementary theoretical studies suggest possible candidates for the structural origin of the defects based on our experimental results. This rapid and scalable nanoindentation method provides a new way to create SPE arrays with specific types of atoms, enabling the comprehensive investigation of the origins and mechanics of SPE formations in two-dimensional (2D) materials and beyond.

The XLZD collaboration is developing a two-phase xenon time projection chamber with an active mass of 60 to 80 t capable of probing the remaining WIMP-nucleon interaction parameter space down to the so-called neutrino fog. In this work we show that, based on the performance of currently operating detectors using the same technology and a realistic reduction of radioactivity in detector materials, such an experiment will also be able to competitively search for neutrinoless double beta decay in $^{136}$Xe using a natural-abundance xenon target. XLZD can reach a 3$\sigma$ discovery potential half-life of 5.7$\times$10$^{27}$ yr (and a 90% CL exclusion of 1.3$\times$10$^{28}$ yr) with 10 years of data taking, corresponding to a Majorana mass range of 7.3-31.3 meV (4.8-20.5 meV). XLZD will thus exclude the inverted neutrino mass ordering parameter space and will start to probe the normal ordering region for most of the nuclear matrix elements commonly considered by the community., Comment: 29 pages, 7 figures

Under the premise that quantum gravity becomes non-negligible, higher-order corrections of non-commutative geometry dominate. In this paper, we studied the thermodynamics of high-order corrections for Schwarzschild-AdS black hole with Lorentz distribution in the framework of non-commutative geometry. Our results indicate that when high-order corrections dominate, the thermodynamic behavior of Schwarzschild-AdS black hole in non-commutative geometry will gradually approach that of ordinary Schwarzschild-AdS black hole. In addition, we also studied the Joule-Thomson effect of Schwarzschild-AdS black hole under high-order corrections.

Real-world noise removal is crucial in low-level computer vision. Due to the remarkable generation capabilities of diffusion models, recent attention has shifted towards leveraging diffusion priors for image restoration tasks. However, existing diffusion priors-based methods either consider simple noise types or rely on approximate posterior estimation, limiting their effectiveness in addressing structured and signal-dependent noise commonly found in real-world images. In this paper, we build upon diffusion priors and propose adaptive likelihood estimation and MAP inference during the reverse diffusion process to tackle real-world noise. We introduce an independent, non-identically distributed likelihood combined with the noise precision (inverse variance) prior and dynamically infer the precision posterior using variational Bayes during the generation process. Meanwhile, we rectify the estimated noise variance through local Gaussian convolution. The final denoised image is obtained by propagating intermediate MAP solutions that balance the updated likelihood and diffusion prior. Additionally, we explore the local diffusion prior inherent in low-resolution diffusion models, enabling direct handling of high-resolution noisy images. Extensive experiments and analyses on diverse real-world datasets demonstrate the effectiveness of our method. Code is available at https://github.com/HUST-Tan/DiffusionVI., Comment: Accepted by NeurIPS2024 as Spotlight

AI holds promise for transforming scientific processes, including hypothesis generation. Prior work on hypothesis generation can be broadly categorized into theory-driven and data-driven approaches. While both have proven effective in generating novel and plausible hypotheses, it remains an open question whether they can complement each other. To address this, we develop the first method that combines literature-based insights with data to perform LLM-powered hypothesis generation. We apply our method on five different datasets and demonstrate that integrating literature and data outperforms other baselines (8.97\% over few-shot, 15.75\% over literature-based alone, and 3.37\% over data-driven alone). Additionally, we conduct the first human evaluation to assess the utility of LLM-generated hypotheses in assisting human decision-making on two challenging tasks: deception detection and AI generated content detection. Our results show that human accuracy improves significantly by 7.44\% and 14.19\% on these tasks, respectively. These findings suggest that integrating literature-based and data-driven approaches provides a comprehensive and nuanced framework for hypothesis generation and could open new avenues for scientific inquiry., Comment: 30 pages, 7 figures, code link: https://github.com/ChicagoHAI/hypothesis-generation

This paper focuses on creating synthetic data to improve the quality of image captions. Existing works typically have two shortcomings. First, they caption images from scratch, ignoring existing alt-text metadata, and second, lack transparency if the captioners' training data (e.g. GPT) is unknown. In this paper, we study a principled approach Altogether based on the key idea to edit and re-align existing alt-texts associated with the images. To generate training data, we perform human annotation where annotators start with the existing alt-text and re-align it to the image content in multiple rounds, consequently constructing captions with rich visual concepts. This differs from prior work that carries out human annotation as a one-time description task solely based on images and annotator knowledge. We train a captioner on this data that generalizes the process of re-aligning alt-texts at scale. Our results show our Altogether approach leads to richer image captions that also improve text-to-image generation and zero-shot image classification tasks., Comment: accepted by EMNLP 2024; MetaCLIPv2

We propose the Large View Synthesis Model (LVSM), a novel transformer-based approach for scalable and generalizable novel view synthesis from sparse-view inputs. We introduce two architectures: (1) an encoder-decoder LVSM, which encodes input image tokens into a fixed number of 1D latent tokens, functioning as a fully learned scene representation, and decodes novel-view images from them; and (2) a decoder-only LVSM, which directly maps input images to novel-view outputs, completely eliminating intermediate scene representations. Both models bypass the 3D inductive biases used in previous methods -- from 3D representations (e.g., NeRF, 3DGS) to network designs (e.g., epipolar projections, plane sweeps) -- addressing novel view synthesis with a fully data-driven approach. While the encoder-decoder model offers faster inference due to its independent latent representation, the decoder-only LVSM achieves superior quality, scalability, and zero-shot generalization, outperforming previous state-of-the-art methods by 1.5 to 3.5 dB PSNR. Comprehensive evaluations across multiple datasets demonstrate that both LVSM variants achieve state-of-the-art novel view synthesis quality. Notably, our models surpass all previous methods even with reduced computational resources (1-2 GPUs). Please see our website for more details: https://haian-jin.github.io/projects/LVSM/ ., Comment: project page: https://haian-jin.github.io/projects/LVSM/

Designing reward functions is a core component of reinforcement learning but can be challenging for truly complex behavior. Reinforcement Learning from Human Feedback (RLHF) has been used to alleviate this challenge by replacing a hand-coded reward function with a reward function learned from preferences. However, it can be exceedingly inefficient to learn these rewards as they are often learned tabula rasa. We investigate whether Large Language Models (LLMs) can reduce this query inefficiency by converting an iterative series of human preferences into code representing the rewards. We propose In-Context Preference Learning (ICPL), a method that uses the grounding of an LLM to accelerate learning reward functions from preferences. ICPL takes the environment context and task description, synthesizes a set of reward functions, and then repeatedly updates the reward functions using human rankings of videos of the resultant policies. Using synthetic preferences, we demonstrate that ICPL is orders of magnitude more efficient than RLHF and is even competitive with methods that use ground-truth reward functions instead of preferences. Finally, we perform a series of human preference-learning trials and observe that ICPL extends beyond synthetic settings and can work effectively with humans-in-the-loop. Additional information and videos are provided at https://sites.google.com/view/few-shot-icpl/home.

Building on the success of large language models (LLMs), recent advancements such as GPT-4o have enabled real-time speech interactions through LLM-based voice assistants, offering a significantly improved user experience compared to traditional text-based interactions. However, the absence of benchmarks designed to evaluate these speech interaction capabilities has hindered progress of LLM-based voice assistants development. Current evaluations focus primarily on automatic speech recognition (ASR) or general knowledge evaluation with clean speeches, neglecting the more intricate, real-world scenarios that involve diverse speaker characteristics, environmental and content factors. To address this, we introduce VoiceBench, the first benchmark designed to provide a multi-faceted evaluation of LLM-based voice assistants. VoiceBench also includes both real and synthetic spoken instructions that incorporate the above three key real-world variations. Extensive experiments reveal the limitations of current LLM-based voice assistant models and offer valuable insights for future research and development in this field., Comment: Work in progress. Data is available at https://github.com/MatthewCYM/VoiceBench

In this study, we present the pulse profile of the unique and the second brightest gamma-ray burst GRB 230307A, and analyze its temporal behavior using a joint GECAM--Fermi/GBM time-resolved spectral analysis. The utilization of GECAM data is advantageous as it successfully captured significant data during the pile-up period of the Fermi/GBM. We investigate the evolution of its flux, photon fluence, photon flux, peak energy, and the corresponding hardness-intensity and hardness-flux correlations. The findings within the first 27 seconds exhibit consistent patterns reported previously, providing valuable insights for comparing observations with predictions from the synchrotron radiation model invoking an expanding shell. Beyond the initial 27 seconds, we observe a notable transition in the emitted radiation, attributed to high latitude emission (HLE), influenced by the geometric properties of the shells and the relativistic Doppler effects. By modeling the data within the framework of the large-radius internal shock model, we discuss the required parameters as well as the limitations of the model. We conclude that a more complicated synchrotron emission model is needed to fully describe the observational data of GRB 230307A., Comment: Accepted for publication in The Astrophysical Journal (ApJ)

This report describes the experimental strategy and technologies for a next-generation xenon observatory sensitive to dark matter and neutrino physics. The detector will have an active liquid xenon target mass of 60-80 tonnes and is proposed by the XENON-LUX-ZEPLIN-DARWIN (XLZD) collaboration. The design is based on the mature liquid xenon time projection chamber technology of the current-generation experiments, LZ and XENONnT. A baseline design and opportunities for further optimization of the individual detector components are discussed. The experiment envisaged here has the capability to explore parameter space for Weakly Interacting Massive Particle (WIMP) dark matter down to the neutrino fog, with a 3$\sigma$ evidence potential for the spin-independent WIMP-nucleon cross sections as low as $3\times10^{-49}\rm cm^2$ (at 40 GeV/c$^2$ WIMP mass). The observatory is also projected to have a 3$\sigma$ observation potential of neutrinoless double-beta decay of $^{136}$Xe at a half-life of up to $5.7\times 10^{27}$ years. Additionally, it is sensitive to astrophysical neutrinos from the atmosphere, sun, and galactic supernovae., Comment: 32 pages, 14 figures

Multimodal Large Language Models (MLLMs) utilize multimodal contexts consisting of text, images, or videos to solve various multimodal tasks. However, we find that changing the order of multimodal input can cause the model's performance to fluctuate between advanced performance and random guessing. This phenomenon exists in both single-modality (text-only or image-only) and mixed-modality (image-text-pair) contexts. Furthermore, we demonstrate that popular MLLMs pay special attention to certain multimodal context positions, particularly the beginning and end. Leveraging this special attention, we place key video frames and important image/text content in special positions within the context and submit them to the MLLM for inference. This method results in average performance gains of 14.7% for video-caption matching and 17.8% for visual question answering tasks. Additionally, we propose a new metric, Position-Invariant Accuracy (PIA), to address order bias in MLLM evaluation. Our research findings contribute to a better understanding of Multi-Modal In-Context Learning (MMICL) and provide practical strategies for enhancing MLLM performance without increasing computational costs.

Stroke-based Rendering (SBR) aims to decompose an input image into a sequence of parameterized strokes, which can be rendered into a painting that resembles the input image. Recently, Neural Painting methods that utilize deep learning and reinforcement learning models to predict the stroke sequences have been developed, but suffer from longer inference time or unstable training. To address these issues, we propose AttentionPainter, an efficient and adaptive model for single-step neural painting. First, we propose a novel scalable stroke predictor, which predicts a large number of stroke parameters within a single forward process, instead of the iterative prediction of previous Reinforcement Learning or auto-regressive methods, which makes AttentionPainter faster than previous neural painting methods. To further increase the training efficiency, we propose a Fast Stroke Stacking algorithm, which brings 13 times acceleration for training. Moreover, we propose Stroke-density Loss, which encourages the model to use small strokes for detailed information, to help improve the reconstruction quality. Finally, we propose a new stroke diffusion model for both conditional and unconditional stroke-based generation, which denoises in the stroke parameter space and facilitates stroke-based inpainting and editing applications helpful for human artists design. Extensive experiments show that AttentionPainter outperforms the state-of-the-art neural painting methods.

Gravitational waves (GWs) offer a new observational window into the universe, providing insights into compact objects and cosmic structures. Gravitational lensing, commonly studied in electromagnetic waves, also affects GWs, introducing magnification, time delays, and multiple images. While existing studies focus on static lenses, many astrophysical lenses are dynamic, with time-varying mass distributions such as moving stars or orbiting binaries. We develop a general theoretical framework to describe non-static lenses and demonstrate how they modulate GW signals, inducing unique time-varying amplitude modulations and spectral broadening. By examining uniformly moving and orbiting binary lenses, we show that these modulations provide new observational signatures, enhancing our understanding of lensing objects and the dynamics of the universe. Our findings have important implications for GW astronomy, offering novel ways to probe lens dynamics and improve the interpretations of GW signals., Comment: 20 pages, 8 figures

We report our experience formally modelling and verifying CXL.cache, the inter-device cache coherence protocol of the Compute Express Link standard. We have used the Isabelle proof assistant to create a formal model for CXL.cache based on the prose English specification. This led to us identifying and proposing fixes to several problems we identified as unclear, ambiguous or inaccurate, some of which could lead to incoherence if left unfixed. Nearly all our issues and proposed fixes have been confirmed and tentatively accepted by the CXL consortium for adoption, save for one which is still under discussion. To validate the faithfulness of our model we performed scenario verification of essential restrictions such as "Snoop-pushes-GO", and produced a fully mechanised proof of a coherence property of the model. The considerable size of this proof, comprising tens of thousands of lemmas, prompted us to develop new proof automation tools, which we have made available for other Isabelle users working with similarly cumbersome proofs., Comment: 12 pages

Due to the rising awareness of privacy protection and the voluminous scale of speech data, it is becoming infeasible for Automatic Speech Recognition (ASR) system developers to train the acoustic model with complete data as before. For example, the data may be owned by different curators, and it is not allowed to share with others. In this paper, we propose a novel paradigm to solve salient problems plaguing the ASR field. In the first stage, multiple acoustic models are trained based upon different subsets of the complete speech data, while in the second phase, two novel algorithms are utilized to generate a high-quality acoustic model based upon those trained on data subsets. We first propose the Genetic Merge Algorithm (GMA), which is a highly specialized algorithm for optimizing acoustic models but suffers from low efficiency. We further propose the SGD-Based Optimizational Merge Algorithm (SOMA), which effectively alleviates the efficiency bottleneck of GMA and maintains superior model accuracy. Extensive experiments on public data show that the proposed methods can significantly outperform the state-of-the-art. Furthermore, we introduce Shapley Value to estimate the contribution score of the trained models, which is useful for evaluating the effectiveness of the data and providing fair incentives to their curators.

Protein structures are important for understanding their functions and interactions. Currently, many protein structure prediction methods are enriching the structure database. Discriminating the origin of structures is crucial for distinguishing between experimentally resolved and computationally predicted structures, evaluating the reliability of prediction methods, and guiding downstream biological studies. Building on works in structure prediction, We developed a structure-sensitive supervised deep learning model, Crystal vs Predicted Evaluator for Protein Structure (CPE-Pro), to represent and discriminate the origin of protein structures. CPE-Pro learns the structural information of proteins and captures inter-structural differences to achieve accurate traceability on four data classes, and is expected to be extended to more. Simultaneously, we utilized Foldseek to encode protein structures into "structure-sequences" and trained a protein Structural Sequence Language Model, SSLM. Preliminary experiments demonstrated that, compared to large-scale protein language models pre-trained on vast amounts of amino acid sequences, the "structure-sequence" enables the language model to learn more informative protein features, enhancing and optimizing structural representations. We have provided the code, model weights, and all related materials on https://github.com/GouWenrui/CPE-Pro-main.git.

Deep neural networks are ubiquitously adopted in many applications, such as computer vision, natural language processing, and graph analytics. However, well-trained neural networks can make prediction errors after deployment as the world changes. \textit{Model editing} involves updating the base model to correct prediction errors with less accessible training data and computational resources. Despite recent advances in model editors in computer vision and natural language processing, editable training in graph neural networks (GNNs) is rarely explored. The challenge with editable GNN training lies in the inherent information aggregation across neighbors, which can lead model editors to affect the predictions of other nodes unintentionally. In this paper, we first observe the gradient of cross-entropy loss for the target node and training nodes with significant inconsistency, which indicates that directly fine-tuning the base model using the loss on the target node deteriorates the performance on training nodes. Motivated by the gradient inconsistency observation, we propose a simple yet effective \underline{G}radient \underline{R}ewiring method for \underline{E}ditable graph neural network training, named \textbf{GRE}. Specifically, we first store the anchor gradient of the loss on training nodes to preserve the locality. Subsequently, we rewire the gradient of the loss on the target node to preserve performance on the training node using anchor gradient. Experiments demonstrate the effectiveness of GRE on various model architectures and graph datasets in terms of multiple editing situations. The source code is available at \url{https://github.com/zhimengj0326/Gradient_rewiring_editing}, Comment: NeurIPS 2024

Concept Bottleneck Models (CBMs) have been proposed as a compromise between white-box and black-box models, aiming to achieve interpretability without sacrificing accuracy. The standard training procedure for CBMs is to predefine a candidate set of human-interpretable concepts, extract their values from the training data, and identify a sparse subset as inputs to a transparent prediction model. However, such approaches are often hampered by the tradeoff between enumerating a sufficiently large set of concepts to include those that are truly relevant versus controlling the cost of obtaining concept extractions. This work investigates a novel approach that sidesteps these challenges: BC-LLM iteratively searches over a potentially infinite set of concepts within a Bayesian framework, in which Large Language Models (LLMs) serve as both a concept extraction mechanism and prior. BC-LLM is broadly applicable and multi-modal. Despite imperfections in LLMs, we prove that BC-LLM can provide rigorous statistical inference and uncertainty quantification. In experiments, it outperforms comparator methods including black-box models, converges more rapidly towards relevant concepts and away from spuriously correlated ones, and is more robust to out-of-distribution samples.

Large generative models (LMs) are increasingly being considered for high-stakes decision-making. This work considers how such models compare to humans and predictive AI models on a specific case of recidivism prediction. We combine three datasets -- COMPAS predictive AI risk scores, human recidivism judgements, and photos -- into a dataset on which we study the properties of several state-of-the-art, multimodal LMs. Beyond accuracy and bias, we focus on studying human-LM alignment on the task of recidivism prediction. We investigate if these models can be steered towards human decisions, the impact of adding photos, and whether anti-discimination prompting is effective. We find that LMs can be steered to outperform humans and COMPAS using in context-learning. We find anti-discrimination prompting to have unintended effects, causing some models to inhibit themselves and significantly reduce their number of positive predictions.

The advancements in cloud-based Large Languages Models (LLMs) have revolutionized AI-assisted programming. However, their integration into certain local development environments like ones within the Apple software ecosystem (e.g., iOS apps, macOS) remains challenging due to computational demands and sandboxed constraints. This paper presents CAMP, a multi-model AI-assisted programming framework that consists of a local model that employs Retrieval-Augmented Generation (RAG) to retrieve contextual information from the codebase to facilitate context-aware prompt construction thus optimizing the performance of the cloud model, empowering LLMs' capabilities in local Integrated Development Environments (IDEs). The methodology is actualized in Copilot for Xcode, an AI-assisted programming tool crafted for Xcode that employs the RAG module to address software constraints and enables diverse generative programming tasks, including automatic code completion, documentation, error detection, and intelligent user-agent interaction. The results from objective experiments on generated code quality and subjective experiments on user adoption collectively demonstrate the pilot success of the proposed system and mark its significant contributions to the realm of AI-assisted programming., Comment: 12 pages, 3 figures, 4 tables

Colorectal cancer is the most common cancer in Singapore and the third most common cancer worldwide. Blood in a person's stool is a symptom of this disease, and it is usually detected by the faecal occult blood test (FOBT). However, the FOBT presents several limitations - the collection process for the stool samples is tedious and unpleasant, the waiting period for results is about 2 weeks and costs are involved. In this research, we propose a simple-to-use, fast and cost-free alternative - a stool recognition neural network that determines if there is blood in one's stool (which indicates a possible risk of colorectal cancer) from an image of it. As this is a new classification task, there was limited data available, hindering classifier performance. Hence, various Generative Adversarial Networks (GANs) (DiffAugment StyleGAN2, DCGAN, Conditional GAN) were trained to generate images of high fidelity to supplement the dataset. Subsequently, images generated by the GAN with the most realistic images (DiffAugment StyleGAN2) were concatenated to the classifier's training batch on-the-fly, improving accuracy to 94%. This model was then deployed to a mobile app - Poolice, where users can take a photo of their stool and obtain instantaneous results if there is blood in their stool, prompting those who do to seek medical advice. As "early detection saves lives", we hope our app built on our stool recognition neural network can help people detect colorectal cancer earlier, so they can seek treatment and have higher chances of survival., Comment: 21 pages, 28 figures

Diagnosing and treating skin diseases require advanced visual skills across multiple domains and the ability to synthesize information from various imaging modalities. Current deep learning models, while effective at specific tasks such as diagnosing skin cancer from dermoscopic images, fall short in addressing the complex, multimodal demands of clinical practice. Here, we introduce PanDerm, a multimodal dermatology foundation model pretrained through self-supervised learning on a dataset of over 2 million real-world images of skin diseases, sourced from 11 clinical institutions across 4 imaging modalities. We evaluated PanDerm on 28 diverse datasets covering a range of clinical tasks, including skin cancer screening, phenotype assessment and risk stratification, diagnosis of neoplastic and inflammatory skin diseases, skin lesion segmentation, change monitoring, and metastasis prediction and prognosis. PanDerm achieved state-of-the-art performance across all evaluated tasks, often outperforming existing models even when using only 5-10% of labeled data. PanDerm's clinical utility was demonstrated through reader studies in real-world clinical settings across multiple imaging modalities. It outperformed clinicians by 10.2% in early-stage melanoma detection accuracy and enhanced clinicians' multiclass skin cancer diagnostic accuracy by 11% in a collaborative human-AI setting. Additionally, PanDerm demonstrated robust performance across diverse demographic factors, including different body locations, age groups, genders, and skin tones. The strong results in benchmark evaluations and real-world clinical scenarios suggest that PanDerm could enhance the management of skin diseases and serve as a model for developing multimodal foundation models in other medical specialties, potentially accelerating the integration of AI support in healthcare., Comment: 56 pages; Technical report

Feature Selection (FS) under domain adaptation (DA) is a critical task in machine learning, especially when dealing with limited target data. However, existing methods lack the capability to guarantee the reliability of FS under DA. In this paper, we introduce a novel statistical method to statistically test FS reliability under DA, named SFS-DA (statistical FS-DA). The key strength of SFS-DA lies in its ability to control the false positive rate (FPR) below a pre-specified level $\alpha$ (e.g., 0.05) while maximizing the true positive rate. Compared to the literature on statistical FS, SFS-DA presents a unique challenge in addressing the effect of DA to ensure the validity of the inference on FS results. We overcome this challenge by leveraging the Selective Inference (SI) framework. Specifically, by carefully examining the FS process under DA whose operations can be characterized by linear and quadratic inequalities, we prove that achieving FPR control in SFS-DA is indeed possible. Furthermore, we enhance the true detection rate by introducing a more strategic approach. Experiments conducted on both synthetic and real-world datasets robustly support our theoretical results, showcasing the superior performance of the proposed SFS-DA method.

The purpose of RGB-D Salient Object Detection (SOD) is to pinpoint the most visually conspicuous areas within images accurately. While conventional deep models heavily rely on CNN extractors and overlook the long-range contextual dependencies, subsequent transformer-based models have addressed the issue to some extent but introduce high computational complexity. Moreover, incorporating spatial information from depth maps has been proven effective for this task. A primary challenge of this issue is how to fuse the complementary information from RGB and depth effectively. In this paper, we propose a dual Mamba-driven cross-modal fusion network for RGB-D SOD, named MambaSOD. Specifically, we first employ a dual Mamba-driven feature extractor for both RGB and depth to model the long-range dependencies in multiple modality inputs with linear complexity. Then, we design a cross-modal fusion Mamba for the captured multi-modal features to fully utilize the complementary information between the RGB and depth features. To the best of our knowledge, this work is the first attempt to explore the potential of the Mamba in the RGB-D SOD task, offering a novel perspective. Numerous experiments conducted on six prevailing datasets demonstrate our method's superiority over sixteen state-of-the-art RGB-D SOD models. The source code will be released at https://github.com/YueZhan721/MambaSOD.

Molecular relational learning (MRL) is crucial for understanding the interaction behaviors between molecular pairs, a critical aspect of drug discovery and development. However, the large feasible model space of MRL poses significant challenges to benchmarking, and existing MRL frameworks face limitations in flexibility and scope. To address these challenges, avoid repetitive coding efforts, and ensure fair comparison of models, we introduce FlexMol, a comprehensive toolkit designed to facilitate the construction and evaluation of diverse model architectures across various datasets and performance metrics. FlexMol offers a robust suite of preset model components, including 16 drug encoders, 13 protein sequence encoders, 9 protein structure encoders, and 7 interaction layers. With its easy-to-use API and flexibility, FlexMol supports the dynamic construction of over 70, 000 distinct combinations of model architectures. Additionally, we provide detailed benchmark results and code examples to demonstrate FlexMol's effectiveness in simplifying and standardizing MRL model development and comparison.

In this study, we explore how the extragalactic background light (EBL) absorption effect influences the photon to axionlike particle (ALP) conversions from the very-high-energy gamma-ray spectral irregularities. For our analysis, we select two well-known BL Lac blazars: Markarian 421 and Markarian 501 with their low and well-defined redshifts $z_0=0.031$ and 0.034, respectively. Their gamma-ray data are recently measured by Fermi-LAT and HAWC with the 1038 days of exposure from 2015 June to 2018 July. We first discuss the EBL absorption effect on the gamma-ray spectral energy distributions by using three common EBL spectral models: Franceschini-08, Finke-10, and Gilmore-12. Then we consider the photon-ALP conversions in the astrophysical magnetic fields. Under the ALP assumption with the parameter space of $\{m_a, g_{a\gamma}\}$, we calculate the best-fit chi-square distribution of the EBL models and define a new delta chi-square $\chi_d^2$ to quantify the chi-square difference. Our results show that the impact from these different EBL spectral models are non-dominated at the low-redshift gamma-ray axionscope., Comment: 17 pages, 5 figures

How embedded, actively accreting low-mass protostars accrete their mass is still greatly debated. Observations are now piecing together the puzzle of embedded protostellar accretion, in particular with new facilities in the near-infrared. However, high-resolution theoretical models are still lacking, with a stark paucity of detailed simulations of these early phases. Here we present high-resolution non-ideal magneto-hydrodynamic simulations of a Solar mass protostar accreting at rates exceeding 10$^{-6} M_{\odot}$ yr$^{-1}$. We show the results of the accretion flow for four different protostellar magnetic fields, 10 G, 500 G, 1 kG, and 2 kG, combined with a disk magnetic field. For weaker (10 G and 500 G) protostar magnetic fields, accretion occurs via a turbulent boundary layer mode, with disk material impacting across the protostellar surface. In the 500 G model, the presence of a magnetically dominated outflow focuses the accretion towards the equator, slightly enhancing and ordering the accretion. For kG magnetic fields, the disk becomes truncated due to the protostellar dipole and exhibits magnetospheric accretion, with the 2 kG model having accretion bursts induced by the interchange instability. We present bolometric light curves for the models and find that they reproduce observations of Class I protostars from YSOVAR, with high bursts followed by an exponential decay possibly being a signature of instability-driven accretion. Finally, we present the filling fractions of accretion and find that 90\% of the mass is accreted in a surface area fraction of 10-20\%. These simulations will be extended in future work for a broader parameter space, with their high resolution and high temporal spacing able to explore a wide range of interesting protostellar physics., Comment: Accepted to A&A

Informal sensemaking surrounding U.S. election processes has been fraught in recent years, due to the inherent uncertainty of elections, the complexity of election processes in the U.S., and to disinformation. Based on insights from qualitative analysis of election rumors spreading online in 2020 and 2022, we introduce the concept of manipulated sensemaking to describe how disinformation functions by disrupting online audiences ability to make sense of novel, uncertain, or ambiguous information. We describe how at the core of this disruption is the ability for disinformation to shape broad, underlying stories called deep stories which determine the frames we use to make sense of this novel information. Additionally, we explain how sensemakings orientation around plausible explanations over accurate explanations makes it vulnerable to manipulation. Lastly, we demonstrate how disinformed deep stories shape sensemaking not just for a single event, but for many events in the future., Comment: 10 pages, CHI 2024 Sensemaking workshop

Sparse Mixture of Experts (SMoE) has become the key to unlocking unparalleled scalability in deep learning. SMoE has the potential to exponentially increase parameter count while maintaining the efficiency of the model by only activating a small subset of these parameters for a given sample. However, it has been observed that SMoE suffers from unstable training and has difficulty adapting to new distributions, leading to the model's lack of robustness to data contamination. To overcome these limitations, we first establish a connection between the dynamics of the expert representations in SMoEs and gradient descent on a multi-objective optimization problem. Leveraging our framework, we then integrate momentum into SMoE and propose a new family of SMoEs named MomentumSMoE. We theoretically prove and numerically demonstrate that MomentumSMoE is more stable and robust than SMoE. In particular, we verify the advantages of MomentumSMoE over SMoE on a variety of practical tasks including ImageNet-1K object recognition and WikiText-103 language modeling. We demonstrate the applicability of MomentumSMoE to many types of SMoE models, including those in the Sparse MoE model for vision (V-MoE) and the Generalist Language Model (GLaM). We also show that other advanced momentum-based optimization methods, such as Adam, can be easily incorporated into the MomentumSMoE framework for designing new SMoE models with even better performance, almost negligible additional computation cost, and simple implementations., Comment: 10 pages in the main text. Published at NeurIPS 2024. The code is available at https://github.com/rachtsy/MomentumSMoE

Large Language Models (LLMs) have achieved impressive results in various tasks but struggle with hallucination problems and lack of relevant knowledge, especially in deep complex reasoning and knowledge-intensive tasks. Knowledge Graphs (KGs), which capture vast amounts of facts in a structured format, offer a reliable source of knowledge for reasoning. However, existing KG-based LLM reasoning methods face challenges like handling multi-hop reasoning, multi-entity questions, and effectively utilizing graph structures. To address these issues, we propose Paths-over-Graph (PoG), a novel method that enhances LLM reasoning by integrating knowledge reasoning paths from KGs, improving the interpretability and faithfulness of LLM outputs. PoG tackles multi-hop and multi-entity questions through a three-phase dynamic multi-hop path exploration, which combines the inherent knowledge of LLMs with factual knowledge from KGs. In order to improve the efficiency, PoG prunes irrelevant information from the graph exploration first and introduces efficient three-step pruning techniques that incorporate graph structures, LLM prompting, and a pre-trained language model (e.g., SBERT) to effectively narrow down the explored candidate paths. This ensures all reasoning paths contain highly relevant information captured from KGs, making the reasoning faithful and interpretable in problem-solving. PoG innovatively utilizes graph structure to prune the irrelevant noise and represents the first method to implement multi-entity deep path detection on KGs for LLM reasoning tasks. Comprehensive experiments on five benchmark KGQA datasets demonstrate PoG outperforms the state-of-the-art method ToG across GPT-3.5-Turbo and GPT-4, achieving an average accuracy improvement of 18.9%. Notably, PoG with GPT-3.5-Turbo surpasses ToG with GPT-4 by up to 23.9%.

Time series forecasting aids decision-making, especially for stakeholders who rely on accurate predictions, making it very important to understand and explain these models to ensure informed decisions. Traditional explainable AI (XAI) methods, which underline feature or temporal importance, often require expert knowledge. In contrast, natural language explanations (NLEs) are more accessible to laypeople. However, evaluating forecast NLEs is difficult due to the complex causal relationships in time series data. To address this, we introduce two new performance metrics based on simulatability, assessing how well a human surrogate can predict model forecasts using the explanations. Experiments show these metrics differentiate good from poor explanations and align with human judgments. Utilizing these metrics, we further evaluate the ability of state-of-the-art large language models (LLMs) to generate explanations for time series data, finding that numerical reasoning, rather than model size, is the main factor influencing explanation quality.

Large Language Models (LLMs) acquire extensive knowledge and remarkable abilities from extensive text corpora, making them powerful tools for various applications. To make LLMs more usable, aligning them with human preferences is essential. Existing alignment techniques, such as Reinforcement Learning from Human Feedback (RLHF) and Direct Preference Optimization (DPO), typically embed predefined preferences directly within the model's parameters. These methods, however, often result in a static alignment that can not account for the diversity of human preferences in practical applications. In response to this challenge, we propose an effective method, \textbf{MetaAlign}, which aims to help LLMs dynamically align with various explicit or implicit preferences specified at inference time. Experimental results show that LLMs optimized on our meticulously constructed MetaAlign Dataset can effectively align with any preferences specified at the inference stage, validating the feasibility of MetaAlign. We hope that our work can provide some insights into the alignment of language models., Comment: 19 pages, 6 figures

Text-to-SQL parsing involves the translation of natural language queries (NLQs) into their corresponding SQL commands. A principal challenge within this domain is the formulation of SQL queries that are not only syntactically correct but also semantically aligned with the natural language input. However, the intrinsic disparity between the NLQ and the SQL poses a significant challenge. In this research, we introduce Keyword Instruction (KeyInst), a novel method designed to enhance SQL formulation by Large Language Models (LLMs). KeyInst essentially provides guidance on pivotal SQL keywords likely to be part of the final query, thus facilitates a smoother SQL query formulation process. We explore two strategies for integrating KeyInst into Text-to-SQL parsing: a pipeline strategy and a single-pass strategy. The former first generates KeyInst for question, which are then used to prompt LLMs. The latter employs a fine-tuned model to concurrently generate KeyInst and SQL in one step. We developed StrucQL, a benchmark specifically designed for the evaluation of SQL formulation. Extensive experiments on StrucQL and other benchmarks demonstrate that KeyInst significantly improves upon the existing Text-to-SQL prompting techniques.

When assisting people in daily tasks, robots need to accurately interpret visual cues and respond effectively in diverse safety-critical situations, such as sharp objects on the floor. In this context, we present M-CoDAL, a multimodal-dialogue system specifically designed for embodied agents to better understand and communicate in safety-critical situations. The system leverages discourse coherence relations to enhance its contextual understanding and communication abilities. To train this system, we introduce a novel clustering-based active learning mechanism that utilizes an external Large Language Model (LLM) to identify informative instances. Our approach is evaluated using a newly created multimodal dataset comprising 1K safety violations extracted from 2K Reddit images. These violations are annotated using a Large Multimodal Model (LMM) and verified by human annotators. Results with this dataset demonstrate that our approach improves resolution of safety situations, user sentiment, as well as safety of the conversation. Next, we deploy our dialogue system on a Hello Robot Stretch robot and conduct a within-subject user study with real-world participants. In the study, participants role-play two safety scenarios with different levels of severity with the robot and receive interventions from our model and a baseline system powered by OpenAI's ChatGPT. The study results corroborate and extend the findings from automated evaluation, showing that our proposed system is more persuasive and competent in a real-world embodied agent setting.

Full-state quantum circuit simulation requires exponentially increased memory size to store the state vector as the number of qubits scales, presenting significant limitations in classical computing systems. Our paper introduces BMQSim, a novel state vector quantum simulation framework that employs lossy compression to address the memory constraints on graphics processing unit (GPU) machines. BMQSim effectively tackles four major challenges for state-vector simulation with compression: frequent compression/decompression, high memory movement overhead, lack of dedicated error control, and unpredictable memory space requirements. Our work proposes an innovative strategy of circuit partitioning to significantly reduce the frequency of compression occurrences. We introduce a pipeline that seamlessly integrates compression with data movement while concealing its overhead. Additionally, BMQSim incorporates the first GPU-based lossy compression technique with point-wise error control. Furthermore, BMQSim features a two-level memory management system, ensuring efficient and stable execution. Our evaluations demonstrate that BMQSim can simulate the same circuit with over 10 times less memory usage on average, achieving fidelity over 0.99 and maintaining comparable simulation time to other state-of-the-art simulators.

Lepton flavour universality in rare $b\rightarrow s$ transitions is tested for the first time using $B_s^0$ meson decays. The measurements are performed using $pp$ collision data collected by the LHCb experiment between 2011 and 2018, corresponding to a total integrated luminosity of 9$\,{\rm fb}^{-1}$. Branching fraction ratios between the $B_s^0 \rightarrow \phi e^+e^-$ and $B_s^0 \rightarrow \phi \mu^+\mu^-$ decays are measured in three regions of dilepton mass squared, $q^2$, with $0.1 < q^2 < 1.1$, $1.1 < q^2 < 6.0$, and $15 < q^2 < 19\,{\rm GeV}^2/c^4$. The results agree with the Standard Model expectation of lepton flavour universality., Comment: All figures and tables, along with machine-readable versions and any supplementary material and additional information, are available at https://lbfence.cern.ch/alcm/public/analysis/full-details/3513/ (LHCb public pages)

The goal of this paper is to accelerate codec-based speech synthesis systems with minimum sacrifice to speech quality. We propose an enhanced inference method that allows for flexible trade-offs between speed and quality during inference without requiring additional training. Our core idea is to predict multiple tokens per inference step of the AR module using multiple prediction heads, resulting in a linear reduction in synthesis time as the number of heads increases. Furthermore, we introduce a novel speculative decoding technique that utilises a Viterbi-based algorithm to select the optimal sequence of generated tokens at each decoding step. In our experiments, we demonstrate that the time required to predict each token is reduced by a factor of 4 to 5 compared to baseline models, with minimal quality trade-off or even improvement in terms of speech intelligibility. Audio samples are available at: multpletokensprediction.github.io/multipletokensprediction.github.io/., Comment: Submitted to IEEE ICASSP 2025

In spite of the great potential of large language models (LLMs) across various tasks, their deployment on resource-constrained devices remains challenging due to their excessive computational and memory demands. Quantization has emerged as an effective solution by storing weights in reduced precision. However, utilizing low precisions (i.e.~2/3-bit) to substantially alleviate the memory-boundedness of LLM decoding, still suffers from prohibitive performance drop. In this work, we argue that existing approaches fail to explore the diversity in computational patterns, redundancy, and sensitivity to approximations of the different phases of LLM inference, resorting to a uniform quantization policy throughout. Instead, we propose a novel phase-aware method that selectively allocates precision during different phases of LLM inference, achieving both strong context extraction during prefill and efficient memory bandwidth utilization during decoding. To further address the memory-boundedness of the decoding phase, we introduce Progressive Mixed-Precision Decoding (PMPD), a technique that enables the gradual lowering of precision deeper in the generated sequence, together with a spectrum of precision-switching schedulers that dynamically drive the precision-lowering decisions in either task-adaptive or prompt-adaptive manner. Extensive evaluation across diverse language tasks shows that when targeting Nvidia GPUs, PMPD achieves 1.4$-$12.2$\times$ speedup in matrix-vector multiplications over fp16 models, while when targeting an LLM-optimized NPU, our approach delivers a throughput gain of 3.8$-$8.0$\times$ over fp16 models and up to 1.54$\times$ over uniform quantization approaches while preserving the output quality.

Large Language Models (LLMs) have demonstrated remarkable performance across various domains, including healthcare. However, their ability to effectively represent structured non-textual data, such as the alphanumeric medical codes used in records like ICD-10 or SNOMED-CT, is limited and has been particularly exposed in recent research. This paper examines the challenges LLMs face in processing medical codes due to the shortcomings of current tokenization methods. As a result, we introduce the UniStruct architecture to design a multimodal medical foundation model of unstructured text and structured data, which addresses these challenges by adapting subword tokenization techniques specifically for the structured medical codes. Our approach is validated through model pre-training on both an extensive internal medical database and a public repository of structured medical records. Trained on over 1 billion tokens on the internal medical database, the proposed model achieves up to a 23% improvement in evaluation metrics, with around 2% gain attributed to our proposed tokenization. Additionally, when evaluated on the EHRSHOT public benchmark with a 1/1000 fraction of the pre-training data, the UniStruct model improves performance on over 42% of the downstream tasks. Our approach not only enhances the representation and generalization capabilities of patient-centric models but also bridges a critical gap in representation learning models' ability to handle complex structured medical data, alongside unstructured text., Comment: NeurIPS 2024 Workshop on Unifying Representations in Neural Models (UniReps 2024)

Challenges in managing linguistic diversity and integrating various musical modalities are faced by current music information retrieval systems. These limitations reduce their effectiveness in a global, multimodal music environment. To address these issues, we introduce CLaMP 2, a system compatible with 101 languages that supports both ABC notation (a text-based musical notation format) and MIDI (Musical Instrument Digital Interface) for music information retrieval. CLaMP 2, pre-trained on 1.5 million ABC-MIDI-text triplets, includes a multilingual text encoder and a multimodal music encoder aligned via contrastive learning. By leveraging large language models, we obtain refined and consistent multilingual descriptions at scale, significantly reducing textual noise and balancing language distribution. Our experiments show that CLaMP 2 achieves state-of-the-art results in both multilingual semantic search and music classification across modalities, thus establishing a new standard for inclusive and global music information retrieval., Comment: 17 pages, 10 figures, 4 tables

The advancement of large language model (LLM) based artificial intelligence technologies has been a game-changer, particularly in sentiment analysis. This progress has enabled a shift from highly specialized research environments to practical, widespread applications within the industry. However, integrating diverse AI models for processing complex multimodal data and the associated high costs of feature extraction presents significant challenges. Motivated by the marketing oriented software development +needs, our study introduces a collaborative AI framework designed to efficiently distribute and resolve tasks across various AI systems to address these issues. Initially, we elucidate the key solutions derived from our development process, highlighting the role of generative AI models like \emph{chatgpt}, \emph{google gemini} in simplifying intricate sentiment analysis tasks into manageable, phased objectives. Furthermore, we present a detailed case study utilizing our collaborative AI system in edge and cloud, showcasing its effectiveness in analyzing sentiments across diverse online media channels.

In this article, we propose a tamed theta Euler-Maruyama (EM) scheme for superlinearly growing neutral McKean-Vlasov stochastic differential delay equations driven by fractional Brownian motion with Hurst exponent $H\in(1/2,1)$. The analytical properties including uniqueness and existence, boundedness of moment and propagation of chaos are investigated. Moreover, the convergence rate of the numerical scheme is established., Comment: arXiv admin note: text overlap with arXiv:2406.09678

Federated Learning (FL) is a privacy-preserving approach that allows servers to aggregate distributed models transmitted from local clients rather than training on user data. More recently, FL has been applied to Speech Emotion Recognition (SER) for secure human-computer interaction applications. Recent research has found that FL is still vulnerable to inference attacks. To this end, this paper focuses on investigating the security of FL for SER concerning property inference attacks. We propose a novel method to protect the property information in speech data by decomposing various properties in the sound and adding perturbations to these properties. Our experiments show that the proposed method offers better privacy-utility trade-offs than existing methods. The trade-offs enable more effective attack prevention while maintaining similar FL utility levels. This work can guide future work on privacy protection methods in speech processing.

Recent advancements in generative AI, exemplified by ChatGPT, have dramatically transformed how people access information. Despite its powerful capabilities, the benefits it provides may not be equally distributed among individuals - a phenomenon referred to as the digital divide. Building upon prior literature, we propose two forms of digital divide in the generative AI adoption process: (i) the learning divide, capturing individuals' heterogeneous abilities to update their perceived utility of ChatGPT; and (ii) the utility divide, representing differences in individuals' actual utility derived from per use of ChatGPT. To evaluate these two divides, we develop a Bayesian learning model that incorporates demographic heterogeneities in both the utility and signal functions. Leveraging a six-month clickstream dataset, we estimate the model and find significant learning and utility divides across various demographic attributes. Interestingly, lower-educated and non-white individuals derive higher utility gains from ChatGPT but learn about its utility at a slower rate. Furthermore, males, younger individuals, and those with an IT background not only derive higher utility per use from ChatGPT but also learn about its utility more rapidly. Besides, we document a phenomenon termed the belief trap, wherein users underestimate ChatGPT's utility, opt not to use the tool, and consequently lack new experiences to update their perceptions, leading to continued underutilization. Our simulation further demonstrates that the learning divide can significantly affect the probability of falling into the belief trap, another form of the digital divide in adoption outcomes (i.e., outcome divide); however, offering training programs can alleviate the belief trap and mitigate the divide., Comment: 43 pages, 3 figures, 6 tables

The application of activity recognition in the ``AI + Education" field is gaining increasing attention. However, current work mainly focuses on the recognition of activities in manually captured videos and a limited number of activity types, with little attention given to recognizing activities in surveillance images from real classrooms. Activity recognition in classroom surveillance images faces multiple challenges, such as class imbalance and high activity similarity. To address this gap, we constructed a novel multimodal dataset focused on classroom surveillance image activity recognition called ARIC (Activity Recognition In Classroom). The ARIC dataset has advantages of multiple perspectives, 32 activity categories, three modalities, and real-world classroom scenarios. In addition to the general activity recognition tasks, we also provide settings for continual learning and few-shot continual learning. We hope that the ARIC dataset can act as a facilitator for future analysis and research for open teaching scenarios. You can download preliminary data from https://ivipclab.github.io/publication_ARIC/ARIC., Comment: arXiv admin note: text overlap with arXiv:2409.03354

Large Language Model (LLM) has revolutionized the field of artificial intelligence, with their capabilities expanding rapidly due to advances in deep learning and increased computational resources. The mixture-of-experts (MoE) model has emerged as a prominent architecture in the field of LLM, better balancing the model performance and computational efficiency. MoE architecture allows for effective scaling and efficient parallel processing, but the GEMM (General Matrix Multiply) of MoE and the large parameters introduce challenges in terms of computation efficiency and communication overhead, which becomes the throughput bottleneck during inference. Applying a single parallelism strategy like EP, DP, PP, etc. to MoE architecture usually achieves sub-optimal inference throughput, the straightforward combinations of existing different parallelisms on MoE can not obtain optimal inference throughput yet. This paper introduces EPS-MoE, a novel expert pipeline scheduler for MoE that goes beyond the existing inference parallelism schemes. Our approach focuses on optimizing the computation of MoE FFN (FeedForward Network) modules by dynamically selecting the best kernel implementation of GroupGemm and DenseGemm for different loads and adaptively overlapping these computations with \textit{all2all} communication, leading to a substantial increase in throughput. Our experimental results demonstrate an average 21% improvement in prefill throughput over existing parallel inference methods. Specifically, we validated our method on DeepSeekV2, a highly optimized model claimed to achieve a prefill throughput of 100K tokens per second. By applying EPS-MoE, we further accelerated it to at least 120K tokens per second., Comment: 13 pages, 14 figures

Best-of-N decoding methods instruct large language models (LLMs) to generate multiple solutions, score each using a scoring function, and select the highest scored as the final answer to mathematical reasoning problems. However, this repeated independent process often leads to the same mistakes, making the selected solution still incorrect. We propose a novel prompting method named Stepwise Correction (StepCo) that helps LLMs identify and revise incorrect steps in their generated reasoning paths. It iterates verification and revision phases that employ a process-supervised verifier. The verify-then-revise process not only improves answer correctness but also reduces token consumption with fewer paths needed to generate. With StepCo, a series of LLMs demonstrate exceptional performance. Notably, using GPT-4o as the backend LLM, StepCo achieves an average accuracy of 94.1 across eight datasets, significantly outperforming the state-of-the-art Best-of-N method by +2.4, while reducing token consumption by 77.8%., Comment: under review