Search

Time series forecasting requires balancing short-term and long-term dependencies for accurate predictions. Existing methods mainly focus on long-term dependency modeling, neglecting the complexities of short-term dynamics, which may hinder performance. Transformers are superior in modeling long-term dependencies but are criticized for their quadratic computational cost. Mamba provides a near-linear alternative but is reported less effective in time series longterm forecasting due to potential information loss. Current architectures fall short in offering both high efficiency and strong performance for long-term dependency modeling. To address these challenges, we introduce Mixture of Universals (MoU), a versatile model to capture both short-term and long-term dependencies for enhancing performance in time series forecasting. MoU is composed of two novel designs: Mixture of Feature Extractors (MoF), an adaptive method designed to improve time series patch representations for short-term dependency, and Mixture of Architectures (MoA), which hierarchically integrates Mamba, FeedForward, Convolution, and Self-Attention architectures in a specialized order to model long-term dependency from a hybrid perspective. The proposed approach achieves state-of-the-art performance while maintaining relatively low computational costs. Extensive experiments on seven real-world datasets demonstrate the superiority of MoU. Code is available at https://github.com/lunaaa95/mou/., Comment: Code at https://github.com/lunaaa95/mou/

Mix-up is a key technique for consistency regularization-based semi-supervised learning methods, generating strong-perturbed samples for strong-weak pseudo-supervision. Existing mix-up operations are performed either randomly or with predefined rules, such as replacing low-confidence patches with high-confidence ones. The former lacks control over the perturbation degree, leading to overfitting on randomly perturbed samples, while the latter tends to generate images with trivial perturbations, both of which limit the effectiveness of consistency learning. This paper aims to answer the following question: How can image mix-up perturbation be adaptively performed during training? To this end, we propose an Adaptive Mix algorithm (AdaMix) for image mix-up in a self-paced learning manner. Given that, in general, a model's performance gradually improves during training, AdaMix is equipped with a self-paced curriculum that, in the initial training stage, provides relatively simple perturbed samples and then gradually increases the difficulty of perturbed images by adaptively controlling the perturbation degree based on the model's learning state estimated by a self-paced regularize. We develop three frameworks with our AdaMix, i.e., AdaMix-ST, AdaMix-MT, and AdaMix-CT, for semi-supervised medical image segmentation. Extensive experiments on three public datasets, including both 2D and 3D modalities, show that the proposed frameworks are capable of achieving superior performance. For example, compared with the state-of-the-art, AdaMix-CT achieves relative improvements of 2.62% in Dice and 48.25% in average surface distance on the ACDC dataset with 10% labeled data. The results demonstrate that mix-up operations with dynamically adjusted perturbation strength based on the segmentation model's state can significantly enhance the effectiveness of consistency regularization.

Recently, there has been a surge of interest in developing graph neural networks that utilize the invariance principle on graphs to generalize the out-of-distribution (OOD) data. Due to the limited knowledge about OOD data, existing approaches often pose assumptions about the correlation strengths of the underlying spurious features and the target labels. However, this prior is often unavailable and will change arbitrarily in the real-world scenarios, which may lead to severe failures of the existing graph invariance learning methods. To bridge this gap, we introduce a novel graph invariance learning paradigm, which induces a robust and general inductive bias. The paradigm is built upon the observation that the infomax principle encourages learning spurious features regardless of spurious correlation strengths. We further propose the EQuAD framework that realizes this learning paradigm and employs tailored learning objectives that provably elicit invariant features by disentangling them from the spurious features learned through infomax. Notably, EQuAD shows stable and enhanced performance across different degrees of bias in synthetic datasets and challenging real-world datasets up to $31.76\%$. Our code is available at \url{https://github.com/tianyao-aka/EQuAD}., Comment: ICML2024 camera-ready version

This work presents a Fully BInarized Large Language Model (FBI-LLM), demonstrating for the first time how to train a large-scale binary language model from scratch (not the partial binary or ternary LLM like BitNet b1.58) to match the performance of its full-precision counterparts (e.g., FP16 or BF16) in transformer-based LLMs. It achieves this by employing an autoregressive distillation (AD) loss with maintaining equivalent model dimensions (130M, 1.3B, 7B) and training data volume as regular LLM pretraining, while delivering competitive results in terms of perplexity and task-specific effectiveness. Intriguingly, by analyzing the training trajectory, we find that the pretrained weight is not necessary for training binarized LLMs from scratch. This research encourages a new computational framework and may facilitate the future design of specialized hardware tailored for fully 1-bit LLMs. We make all models, code, and training dataset fully accessible and transparent to support further research (Code: https://github.com/LiqunMa/FBI-LLM. Model: https://huggingface.co/LiqunMa/)., Comment: Github at https://github.com/LiqunMa/FBI-LLM

The existing barely-supervised medical image segmentation (BSS) methods, adopting a registration-segmentation paradigm, aim to learn from data with very few annotations to mitigate the extreme label scarcity problem. However, this paradigm poses a challenge: pseudo-labels generated by image registration come with significant noise. To address this issue, we propose a self-paced sample selection framework (SPSS) for BSS. Specifically, SPSS comprises two main components: 1) self-paced uncertainty sample selection (SU) for explicitly improving the quality of pseudo labels in the image space, and 2) self-paced bidirectional feature contrastive learning (SC) for implicitly improving the quality of pseudo labels through enhancing the separability between class semantics in the feature space. Both SU and SC are trained collaboratively in a self-paced learning manner, ensuring that SPSS can learn from high-quality pseudo labels for BSS. Extensive experiments on two public medical image segmentation datasets demonstrate the effectiveness and superiority of SPSS over the state-of-the-art. Our code is release at https://github.com/SuuuJM/SPSS., Comment: Accepted to MICCAI 2024

Multimodal large language models (MLLMs) have shown impressive success across modalities such as image, video, and audio in a variety of understanding and generation tasks. However, current MLLMs are surprisingly poor at understanding webpage screenshots and generating their corresponding HTML code. To address this problem, we propose Web2Code, a benchmark consisting of a new large-scale webpage-to-code dataset for instruction tuning and an evaluation framework for the webpage understanding and HTML code translation abilities of MLLMs. For dataset construction, we leverage pretrained LLMs to enhance existing webpage-to-code datasets as well as generate a diverse pool of new webpages rendered into images. Specifically, the inputs are webpage images and instructions, while the responses are the webpage's HTML code. We further include diverse natural language QA pairs about the webpage content in the responses to enable a more comprehensive understanding of the web content. To evaluate model performance in these tasks, we develop an evaluation framework for testing MLLMs' abilities in webpage understanding and web-to-code generation. Extensive experiments show that our proposed dataset is beneficial not only to our proposed tasks but also in the general visual domain, while previous datasets result in worse performance. We hope our work will contribute to the development of general MLLMs suitable for web-based content generation and task automation. Our data and code will be available at https://github.com/MBZUAI-LLM/web2code., Comment: Website at https://mbzuai-llm.github.io/webpage2code/

Multiple-choice questions (MCQ) are frequently used to assess large language models (LLMs). Typically, an LLM is given a question and selects the answer deemed most probable after adjustments for factors like length. Unfortunately, LLMs may inherently favor certain answer choice IDs, such as A/B/C/D, due to inherent biases of priori unbalanced probabilities, influencing the prediction of answers based on these IDs. Previous research has introduced methods to reduce this ''selection bias'' by simply permutating options on a few test samples and applying to new ones. Another problem of MCQ is the lottery ticket choice by ''random guessing''. The LLM does not learn particular knowledge, but the option is guessed correctly. This situation is especially serious for those small-scale LLMs. To address them, a more thorough approach involves shifting from MCQ to open-style questions, which can fundamentally eliminate selection bias and random guessing issues. However, transitioning causes its own set of challenges in (1) identifying suitable open-style questions and (2) validating the correctness of LLM open-style responses against human-annotated ground-truths. This work aims to tackle these significant difficulties, and establish a new LLM evaluation benchmark through entirely open-style questions. Consequently, we introduce the Open-LLM-Leaderboard to track various LLMs' performance and reflect true capability of them, such as GPT-4o/4/3.5, Claude 3, Gemini, etc. Our code and dataset are available at https://github.com/VILA-Lab/Open-LLM-Leaderboard., Comment: Code and dataset are available at https://github.com/VILA-Lab/Open-LLM-Leaderboard

Face Recognition Systems (FRS) have increasingly integrated into critical applications, including surveillance and user authentication, highlighting their pivotal role in modern security systems. Recent studies have revealed vulnerabilities in FRS to adversarial (e.g., adversarial patch attacks) and backdoor attacks (e.g., training data poisoning), raising significant concerns about their reliability and trustworthiness. Previous studies primarily focus on traditional adversarial or backdoor attacks, overlooking the resource-intensive or privileged-manipulation nature of such threats, thus limiting their practical generalization, stealthiness, universality and robustness. Correspondingly, in this paper, we delve into the inherent vulnerabilities in FRS through user studies and preliminary explorations. By exploiting these vulnerabilities, we identify a novel attack, facial identity backdoor attack dubbed FIBA, which unveils a potentially more devastating threat against FRS:an enrollment-stage backdoor attack. FIBA circumvents the limitations of traditional attacks, enabling broad-scale disruption by allowing any attacker donning a specific trigger to bypass these systems. This implies that after a single, poisoned example is inserted into the database, the corresponding trigger becomes a universal key for any attackers to spoof the FRS. This strategy essentially challenges the conventional attacks by initiating at the enrollment stage, dramatically transforming the threat landscape by poisoning the feature database rather than the training data., Comment: 19 pages,version 3

Recent research indicates that large language models (LLMs) are susceptible to jailbreaking attacks that can generate harmful content. This paper introduces a novel token-level attack method, Adaptive Dense-to-Sparse Constrained Optimization (ADC), which effectively jailbreaks several open-source LLMs. Our approach relaxes the discrete jailbreak optimization into a continuous optimization and progressively increases the sparsity of the optimizing vectors. Consequently, our method effectively bridges the gap between discrete and continuous space optimization. Experimental results demonstrate that our method is more effective and efficient than existing token-level methods. On Harmbench, our method achieves state of the art attack success rate on seven out of eight LLMs. Code will be made available. Trigger Warning: This paper contains model behavior that can be offensive in nature.

This paper investigates an extremely challenging problem: barely-supervised volumetric medical image segmentation (BSS). A BSS training dataset consists of two parts: 1) a barely-annotated labeled set, where each labeled image contains only a single-slice annotation, and 2) an unlabeled set comprising numerous unlabeled volumetric images. State-of-the-art BSS methods employ a registration-based paradigm, which uses inter-slice image registration to propagate single-slice annotations into volumetric pseudo labels, constructing a completely annotated labeled set, to which a semi-supervised segmentation scheme can be applied. However, the paradigm has a critical limitation: the pseudo-labels generated by image registration are unreliable and noisy. Motivated by this, we propose a new perspective: instead of solving BSS within a semi-supervised learning scheme, this work formulates BSS as an unsupervised domain adaptation problem. To this end, we propose a novel BSS framework, \textbf{B}arely-supervised learning \textbf{via} unsupervised domain \textbf{A}daptation (BvA), as an alternative to the dominant registration paradigm. Specifically, we first design a novel noise-free labeled data construction algorithm (NFC) for slice-to-volume labeled data synthesis. Then, we introduce a frequency and spatial Mix-Up strategy (FSX) to mitigate the domain shifts. Extensive experiments demonstrate that our method provides a promising alternative for BSS. Remarkably, the proposed method, trained on the left atrial segmentation dataset with \textbf{only one} barely-labeled image, achieves a Dice score of 81.20%, outperforming the state-of-the-art by 61.71%. The code is available at https://github.com/Senyh/BvA.

Dataset condensation, a concept within data-centric learning, efficiently transfers critical attributes from an original dataset to a synthetic version, maintaining both diversity and realism. This approach significantly improves model training efficiency and is adaptable across multiple application areas. Previous methods in dataset condensation have faced challenges: some incur high computational costs which limit scalability to larger datasets (e.g., MTT, DREAM, and TESLA), while others are restricted to less optimal design spaces, which could hinder potential improvements, especially in smaller datasets (e.g., SRe2L, G-VBSM, and RDED). To address these limitations, we propose a comprehensive design framework that includes specific, effective strategies like implementing soft category-aware matching and adjusting the learning rate schedule. These strategies are grounded in empirical evidence and theoretical backing. Our resulting approach, Elucidate Dataset Condensation (EDC), establishes a benchmark for both small and large-scale dataset condensation. In our testing, EDC achieves state-of-the-art accuracy, reaching 48.6% on ImageNet-1k with a ResNet-18 model at an IPC of 10, which corresponds to a compression ratio of 0.78%. This performance exceeds those of SRe2L, G-VBSM, and RDED by margins of 27.3%, 17.2%, and 6.6%, respectively.

Proprietary large language models (LLMs) have been widely applied in various scenarios. Additionally, deploying LLMs on edge devices is trending for efficiency and privacy reasons. However, edge deployment of proprietary LLMs introduces new security challenges: edge-deployed models are exposed as white-box accessible to users, enabling adversaries to conduct effective model stealing (MS) attacks. Unfortunately, existing defense mechanisms fail to provide effective protection. Specifically, we identify four critical protection properties that existing methods fail to simultaneously satisfy: (1) maintaining protection after a model is physically copied; (2) authorizing model access at request level; (3) safeguarding runtime reverse engineering; (4) achieving high security with negligible runtime overhead. To address the above issues, we propose TransLinkGuard, a plug-and-play model protection approach against model stealing on edge devices. The core part of TransLinkGuard is a lightweight authorization module residing in a secure environment, e.g., TEE. The authorization module can freshly authorize each request based on its input. Extensive experiments show that TransLinkGuard achieves the same security protection as the black-box security guarantees with negligible overhead., Comment: arXiv admin note: text overlap with arXiv:2310.07152 by other authors

Dataset distillation aims to compress information from a large-scale original dataset to a new compact dataset while striving to preserve the utmost degree of the original data informational essence. Previous studies have predominantly concentrated on aligning the intermediate statistics between the original and distilled data, such as weight trajectory, features, gradient, BatchNorm, etc. In this work, we consider addressing this task through the new lens of model informativeness in the compression stage on the original dataset pretraining. We observe that with the prior state-of-the-art SRe$^2$L, as model sizes increase, it becomes increasingly challenging for supervised pretrained models to recover learned information during data synthesis, as the channel-wise mean and variance inside the model are flatting and less informative. We further notice that larger variances in BN statistics from self-supervised models enable larger loss signals to update the recovered data by gradients, enjoying more informativeness during synthesis. Building on this observation, we introduce SC-DD, a simple yet effective Self-supervised Compression framework for Dataset Distillation that facilitates diverse information compression and recovery compared to traditional supervised learning schemes, further reaps the potential of large pretrained models with enhanced capabilities. Extensive experiments are conducted on CIFAR-100, Tiny-ImageNet and ImageNet-1K datasets to demonstrate the superiority of our proposed approach. The proposed SC-DD outperforms all previous state-of-the-art supervised dataset distillation methods when employing larger models, such as SRe$^2$L, MTT, TESLA, DC, CAFE, etc., by large margins under the same recovery and post-training budgets. Code is available at https://github.com/VILA-Lab/SRe2L/tree/main/SCDD/.

We present a new 3D point-based detector model, named Shift-SSD, for precise 3D object detection in autonomous driving. Traditional point-based 3D object detectors often employ architectures that rely on a progressive downsampling of points. While this method effectively reduces computational demands and increases receptive fields, it will compromise the preservation of crucial non-local information for accurate 3D object detection, especially in the complex driving scenarios. To address this, we introduce an intriguing Cross-Cluster Shifting operation to unleash the representation capacity of the point-based detector by efficiently modeling longer-range inter-dependency while including only a negligible overhead. Concretely, the Cross-Cluster Shifting operation enhances the conventional design by shifting partial channels from neighboring clusters, which enables richer interaction with non-local regions and thus enlarges the receptive field of clusters. We conduct extensive experiments on the KITTI, Waymo, and nuScenes datasets, and the results demonstrate the state-of-the-art performance of Shift-SSD in both detection accuracy and runtime efficiency., Comment: ICRA2024

We present the results of our observations using the Giant Meterwave Radio Telescope (GMRT) to investigate the radio continuum and OH line emission of 10 OHM candidates from the Arecibo Legacy Fast ALFA (ALFALFA) survey. Among these candidates, we have identified two sources, AGC115713 and AGC249507, which display compact OH line emission that are spatially associated with radio continuum emission. These characteristics align with the typical properties of OHM galaxies. Furthermore, the infrared (IR) properties of these two galaxies are consistent with those of known OHM galaxies. %Importantly, these two sources have been independently confirmed by alternative methods. Of the two sources, AGC 249507 has been confirmed through optical redshift, whereas AGC 115713 meets a WISE color selection criterion in the literature, providing additional support for this source being an OHM galaxy rather than a nearby \HI galaxy. On the contrary, no significant spectral line emission were detected in the remaining eight OHM candidates using our full GMRT dataset. This suggests that the spectral line emission initially detected by the ALFALFA survey may have been significantly resolved in our high-resolution observations. Additionally, the absence of radio continuum emission in 6 candidates also distinguishes them from known OHM galaxies documented in the literature. These findings support the notion that OHM emission may be distributed on a subarcsecond scale, underscoring the utility of arcsecond-scale observations in confirming OHM candidates, particularly those lacking optical redshift data., Comment: Accepted by A&A

In this paper, we propose a novel knowledge transfer framework that introduces continuous normalizing flows for progressive knowledge transformation and leverages multi-step sampling strategies to achieve precision knowledge transfer. We name this framework Knowledge Transfer with Flow Matching (FM-KT), which can be integrated with a metric-based distillation method with any form (\textit{e.g.} vanilla KD, DKD, PKD and DIST) and a meta-encoder with any available architecture (\textit{e.g.} CNN, MLP and Transformer). By introducing stochastic interpolants, FM-KD is readily amenable to arbitrary noise schedules (\textit{e.g.}, VP-ODE, VE-ODE, Rectified flow) for normalized flow path estimation. We theoretically demonstrate that the training objective of FM-KT is equivalent to minimizing the upper bound of the teacher feature map or logit negative log-likelihood. Besides, FM-KT can be viewed as a unique implicit ensemble method that leads to performance gains. By slightly modifying the FM-KT framework, FM-KT can also be transformed into an online distillation framework OFM-KT with desirable performance gains. Through extensive experiments on CIFAR-100, ImageNet-1k, and MS-COCO datasets, we empirically validate the scalability and state-of-the-art performance of our proposed methods among relevant comparison approaches.

We present FerKD, a novel efficient knowledge distillation framework that incorporates partial soft-hard label adaptation coupled with a region-calibration mechanism. Our approach stems from the observation and intuition that standard data augmentations, such as RandomResizedCrop, tend to transform inputs into diverse conditions: easy positives, hard positives, or hard negatives. In traditional distillation frameworks, these transformed samples are utilized equally through their predictive probabilities derived from pretrained teacher models. However, merely relying on prediction values from a pretrained teacher, a common practice in prior studies, neglects the reliability of these soft label predictions. To address this, we propose a new scheme that calibrates the less-confident regions to be the context using softened hard groundtruth labels. Our approach involves the processes of hard regions mining + calibration. We demonstrate empirically that this method can dramatically improve the convergence speed and final accuracy. Additionally, we find that a consistent mixing strategy can stabilize the distributions of soft supervision, taking advantage of the soft labels. As a result, we introduce a stabilized SelfMix augmentation that weakens the variation of the mixed images and corresponding soft labels through mixing similar regions within the same image. FerKD is an intuitive and well-designed learning system that eliminates several heuristics and hyperparameters in former FKD solution. More importantly, it achieves remarkable improvement on ImageNet-1K and downstream tasks. For instance, FerKD achieves 81.2% on ImageNet-1K with ResNet-50, outperforming FKD and FunMatch by remarkable margins. Leveraging better pre-trained weights and larger architectures, our finetuned ViT-G14 even achieves 89.9%. Our code is available at https://github.com/szq0214/FKD/tree/main/FerKD., Comment: ICCV 2023. Github: https://github.com/szq0214/FKD/tree/main/FerKD

This paper introduces 26 guiding principles designed to streamline the process of querying and prompting large language models. Our goal is to simplify the underlying concepts of formulating questions for various scales of large language models, examining their abilities, and enhancing user comprehension on the behaviors of different scales of large language models when feeding into different prompts. Extensive experiments are conducted on LLaMA-1/2 (7B, 13B and 70B), GPT-3.5/4 to verify the effectiveness of the proposed principles on instructions and prompts design. We hope that this work can provide a better guide for researchers working on the prompting of large language models. Project page is available at https://github.com/VILA-Lab/ATLAS., Comment: Github at: https://github.com/VILA-Lab/ATLAS

The recent surge in open-source Large Language Models (LLMs), such as LLaMA, Falcon, and Mistral, provides diverse options for AI practitioners and researchers. However, most LLMs have only released partial artifacts, such as the final model weights or inference code, and technical reports increasingly limit their scope to high-level design choices and surface statistics. These choices hinder progress in the field by degrading transparency into the training of LLMs and forcing teams to rediscover many details in the training process. We present LLM360, an initiative to fully open-source LLMs, which advocates for all training code and data, model checkpoints, and intermediate results to be made available to the community. The goal of LLM360 is to support open and collaborative AI research by making the end-to-end LLM training process transparent and reproducible by everyone. As a first step of LLM360, we release two 7B parameter LLMs pre-trained from scratch, Amber and CrystalCoder, including their training code, data, intermediate checkpoints, and analyses (at https://www.llm360.ai). We are committed to continually pushing the boundaries of LLMs through this open-source effort. More large-scale and stronger models are underway and will be released in the future.

Magnetic fields of molecular clouds in the Central Molecular Zone (CMZ) have been relatively underobserved at sub-parsec resolution. Here we report JCMT/POL2 observations of polarized dust emission in the CMZ, which reveal magnetic field structures in dense gas at ~0.5 pc resolution. The eleven molecular clouds in our sample including two in the western part of the CMZ (Sgr C and a far-side cloud candidate), four around the Galactic longitude 0 (the 50 km s-1 cloud, CO0.02-0.02, the `Stone' and the `Sticks & Straw' among the Three Little Pigs), and five along the Dust Ridge (G0.253+0.016, clouds b, c, d, and e/f), for each of which we estimate the magnetic field strength using the angular dispersion function method. The morphologies of magnetic fields in the clouds suggest potential imprints of feedback from expanding H II regions and young massive star clusters. A moderate correlation between the total viral parameter versus the star formation rate and the dense gas fraction of the clouds is found. A weak correlation between the mass-to-flux ratio and the star formation rate, and a weak anti-correlation between the magnetic field and the dense gas fraction are also found. Comparisons between magnetic fields and other dynamic components in clouds suggest a more dominant role of self-gravity and turbulence in determining the dynamical states of the clouds and affecting star formation at the studied scales., Comment: ApJ accepted. 26 pages, 13 figures, 5 appendices. Magnetic field segment catalogs are publicly available at https://zenodo.org/doi/10.5281/zenodo.8409806

Studying a centi-parsec supermassive black hole binary (SMBHB) would allow us to explore a new parameter space in active galactic nuclei, and these objects are also potential sources of gravitational waves. We report evidence that an SMBHB with an orbital period of about 30 yr may be resident in the nearby galactic nucleus M81. This orbital period and the known mass of M81 imply an orbital separation of about 0.02 pc. The jet emanating from the primary black hole showed a short period of jet wobbling at about 16.7 yr, superposing a long-term precession at a timescale of several hundred years. Periodic radio and X-ray outbursts were also found two times per orbital period, which could be explained by a double-peaked mass accretion rate variation per binary orbit. If confirmed, M81 would be one of the closest SMBHB candidates, providing a rare opportunity to study the final parsec problem., Comment: 10 pages, 4 figures, published in the ApJ

The lightweight "local-match-global" matching introduced by SRe2L successfully creates a distilled dataset with comprehensive information on the full 224x224 ImageNet-1k. However, this one-sided approach is limited to a particular backbone, layer, and statistics, which limits the improvement of the generalization of a distilled dataset. We suggest that sufficient and various "local-match-global" matching are more precise and effective than a single one and has the ability to create a distilled dataset with richer information and better generalization. We call this perspective "generalized matching" and propose Generalized Various Backbone and Statistical Matching (G-VBSM) in this work, which aims to create a synthetic dataset with densities, ensuring consistency with the complete dataset across various backbones, layers, and statistics. As experimentally demonstrated, G-VBSM is the first algorithm to obtain strong performance across both small-scale and large-scale datasets. Specifically, G-VBSM achieves a performance of 38.7% on CIFAR-100 with 128-width ConvNet, 47.6% on Tiny-ImageNet with ResNet18, and 31.4% on the full 224x224 ImageNet-1k with ResNet18, under images per class (IPC) 10, 50, and 10, respectively. These results surpass all SOTA methods by margins of 3.9%, 6.5%, and 10.1%, respectively., Comment: Accepted by CVPR2024

We conducted a Ka-band (26.1--35 GHz) line survey towards Orion KL using the TianMa 65-m Radio Telescope (TMRT). It is the first blind line survey in the Ka band, and achieves a sensitivity of mK level (1--3 mK at a spectral resolution of $\sim$1 km s$^{-1}$). In total, 592 Gaussian features are extracted. Among them, 257 radio recombination lines (RRLs) are identified. The maximum $\Delta n$ of RRLs of H, He and C are 20, 15, and 5, respectively. Through stacking, we have detected the $\beta$ lines of ion RRLs (RRLs of C$^+$ with possible contribution of other ions like O$^+$) for the first time, and tentative signal of the $\gamma$ lines of ion RRLs can also be seen on the stacked spectrum. Besides, 318 other line features were assigned to 37 molecular species, and ten of these species were not detected in the Q-band survey of TMRT. The vibrationally excited states of nine species were also detected. Emission of most species can be modeled under LTE. A number of transitions of E-CH3OH ($J_2-J_1$) display maser effects, which are confirmed by our modeling, and besides the bumping peak at $J\sim 6$ there is another peak at $J\sim 13$. Methylcyanoacetylene (CH$_3$C$_3$N) is detected in Orion KL for the first time. This work emphasizes that the Ka band, which was long-ignored for spectral line surveys, is very useful for surveying RRLs and molecular lines simultaneously., Comment: accepted by ApJS

Modern computer vision offers a great variety of models to practitioners, and selecting a model from multiple options for specific applications can be challenging. Conventionally, competing model architectures and training protocols are compared by their classification accuracy on ImageNet. However, this single metric does not fully capture performance nuances critical for specialized tasks. In this work, we conduct an in-depth comparative analysis of model behaviors beyond ImageNet accuracy, for both ConvNet and Vision Transformer architectures, each across supervised and CLIP training paradigms. Although our selected models have similar ImageNet accuracies and compute requirements, we find that they differ in many other aspects: types of mistakes, output calibration, transferability, and feature invariance, among others. This diversity in model characteristics, not captured by traditional metrics, highlights the need for more nuanced analysis when choosing among different models. Our code is available at https://github.com/kirill-vish/Beyond-INet., Comment: Project page: https://kirill-vish.github.io/beyond-imagenet-accuracy/

This paper presents an overview of the QUARKS survey, which stands for `Querying Underlying mechanisms of massive star formation with ALMA-Resolved gas Kinematics and Structures'. The QUARKS survey is observing 139 massive clumps covered by 156 pointings at ALMA Band 6 ($\lambda\sim$ 1.3 mm). In conjunction with data obtained from the ALMA-ATOMS survey at Band 3 ($\lambda\sim$ 3 mm), QUARKS aims to carry out an unbiased statistical investigation of massive star formation process within protoclusters down to a scale of 1000 au. This overview paper describes the observations and data reduction of the QUARKS survey, and gives a first look at an exemplar source, the mini-starburst Sgr B2(M). The wide-bandwidth (7.5 GHz) and high-angular-resolution (~0.3 arcsec) observations of the QUARKS survey allow to resolve much more compact cores than could be done by the ATOMS survey, and to detect previously unrevealed fainter filamentary structures. The spectral windows cover transitions of species including CO, SO, N$_2$D$^+$, SiO, H$_{30}\alpha$, H$_2$CO, CH$_3$CN and many other complex organic molecules, tracing gas components with different temperatures and spatial extents. QUARKS aims to deepen our understanding of several scientific topics of massive star formation, such as the mass transport within protoclusters by (hub-)filamentary structures, the existence of massive starless cores, the physical and chemical properties of dense cores within protoclusters, and the feedback from already formed high-mass young protostars., Comment: 9 figures, 4 tables, accepted by RAA

Large Language Models (LLMs) with billions of parameters are prime targets for network pruning, removing some model weights without hurting performance. Prior approaches such as magnitude pruning, SparseGPT, and Wanda, either concentrated solely on weights or integrated weights with activations for sparsity. However, they overlooked the informative gradients derived from pretrained LLMs. In this paper, we present a novel sparsity-centric pruning method for pretrained LLMs, termed Gradient-based Language Model Pruner (GBLM-Pruner). GBLM-Pruner leverages the first-order term of the Taylor expansion, operating in a training-free manner by harnessing properly normalized gradients from a few calibration samples to determine the pruning metric, and substantially outperforms competitive counterparts like SparseGPT and Wanda in multiple benchmarks. Intriguingly, by incorporating gradients, unstructured pruning with our method tends to reveal some structural patterns, which mirrors the geometric interdependence inherent in the LLMs' parameter structure. Additionally, GBLM-Pruner functions without any subsequent retraining or weight updates to maintain its simplicity as other counterparts. Extensive evaluations on LLaMA-1 and LLaMA-2 across various benchmarks show that GBLM-Pruner surpasses magnitude pruning, Wanda and SparseGPT by significant margins. We further extend our approach on Vision Transformer. Our code and models are available at https://github.com/VILA-Lab/GBLM-Pruner., Comment: Code and models at https://github.com/VILA-Lab/GBLM-Pruner

The peptide-like molecules, cyanoformamide (NCCONH2), is the cyano (CN) derivative of formamide (NH2CHO). It is known to play a role in the synthesis of nucleic acid precursors under prebiotic conditions. In this paper, we present a tentative detection of NCCONH2 in the interstellar medium (ISM) with the Atacama Large Millimeter/submillimeter Array (ALMA) archive data. Ten unblended lines of NCCONH2 were seen around 3sigma noise levels toward Sagittarius B2(N1E), a position that is slightly offset from the continuum peak. The column density of NCCONH2 was estimated to be 2.4\times 10^15 cm ^-2, and the fractional abundance of NCCONH2 toward Sgr B2(N1E) was 6.9\times10^-10. The abundance ratio between NCCONH2 and NH2CHO is estimated to be ~0.01. We also searched for other peptide-like molecules toward Sgr B2(N1E). The abundances of NH2CHO, CH3NCO and CH3NHCHO toward Sgr B2(N1E) were about one tenth of those toward Sgr B2(N1S), while the abundances of CH3CONH2 was only one twentieth of that toward Sgr B2(N1S)., Comment: 20 pages, 6 figures, 2 tables, accepted by PASJ

The nearby radio galaxy M87 offers a unique opportunity to explore the connections between the central supermassive black hole and relativistic jets. Previous studies of the inner region of M87 revealed a wide opening angle for the jet originating near the black hole. The Event Horizon Telescope resolved the central radio source and found an asymmetric ring structure consistent with expectations from General Relativity. With a baseline of 17 years of observations, there was a shift in the jet's transverse position, possibly arising from an eight to ten-year quasi-periodicity. However, the origin of this sideways shift remains unclear. Here we report an analysis of radio observations over 22 years that suggests a period of about 11 years in the position angle variation of the jet. We infer that we are seeing a spinning black hole that induces the Lense-Thirring precession of a misaligned accretion disk. Similar jet precession may commonly occur in other active galactic nuclei but has been challenging to detect owing to the small magnitude and long period of the variation., Comment: 41 pages, 7 figures, 7 tables

Through very long baseline interferometry observations of one of the closest low-luminosity active galactic nuclei M81* at multifrequencies of 8.8, 22 and 44GHz, a bright discrete knot with an unusual low apparent speed $\sim$0.1c was detected. Combining with the contemporary monitoring of X-rays data at 2-10keV, it indicates that a moderate X-ray flare happened when the knot launched from the core region. Three possible origins of the knot are proposed to explain our observational results. They are an episodic jet ejection, a low-speed shock wave, and a possible secondary black hole in a binary system, respectively. Future intensive multiwavelength monitoring can help to understand the discrete knot as well as the central black hole better.

This paper aims to understand the impacts of various data combinations (e.g., web text, Wikipedia, GitHub, books) on the pretraining of large language models using SlimPajama. SlimPajama is a rigorously deduplicated, multi-source dataset, which has been refined and further deduplicated to 627B tokens from the extensive 1.2T token RedPajama dataset contributed by Together. We have termed our research as SlimPajama-DC, an empirical analysis designed to uncover fundamental characteristics and best practices associated with employing SlimPajama in the training of large language models. During our research with SlimPajama, two pivotal observations emerged: (1) Global deduplication vs. local deduplication. We analyze and discuss how global (across different sources of datasets) and local (within the single source of dataset) deduplications affect the performance of trained models. (2) Proportions of highly-deduplicated multi-source datasets in the combination. To study this, we construct six configurations on SlimPajama dataset and train individual ones using 1.3B Cerebras-GPT model with Alibi and SwiGLU. Our best configuration outperforms the 1.3B model trained on RedPajama using the same number of training tokens by a significant margin. All our 1.3B models are trained on Cerebras 16$\times$ CS-2 cluster with a total of 80 PFLOP/s in bf16 mixed precision. We further extend our discoveries (such as increasing data diversity is crucial after global deduplication) on a 7B model with large batch-size training. Our SlimPajama-DC models are available at: https://huggingface.co/MBZUAI-LLM/SlimPajama-DC and the separate SlimPajama-DC datasets are available at: https://huggingface.co/datasets/MBZUAI-LLM/SlimPajama-627B-DC., Comment: Technical report. Models at: https://huggingface.co/MBZUAI-LLM/SlimPajama-DC and dataset at: https://huggingface.co/datasets/MBZUAI-LLM/SlimPajama-627B-DC

We introduce Jais and Jais-chat, new state-of-the-art Arabic-centric foundation and instruction-tuned open generative large language models (LLMs). The models are based on the GPT-3 decoder-only architecture and are pretrained on a mixture of Arabic and English texts, including source code in various programming languages. With 13 billion parameters, they demonstrate better knowledge and reasoning capabilities in Arabic than any existing open Arabic and multilingual models by a sizable margin, based on extensive evaluation. Moreover, the models are competitive in English compared to English-centric open models of similar size, despite being trained on much less English data. We provide a detailed description of the training, the tuning, the safety alignment, and the evaluation of the models. We release two open versions of the model -- the foundation Jais model, and an instruction-tuned Jais-chat variant -- with the aim of promoting research on Arabic LLMs. Available at https://huggingface.co/inception-mbzuai/jais-13b-chat, Comment: Arabic-centric, foundation model, large-language model, LLM, generative model, instruction-tuned, Jais, Jais-chat

The Event Horizon Telescope (EHT) observed in 2017 the supermassive black hole at the center of the Milky Way, Sagittarius A* (Sgr A*), at a frequency of 228.1 GHz ($\lambda$=1.3 mm). The fundamental physics tests that even a single pulsar orbiting Sgr A* would enable motivate searching for pulsars in EHT datasets. The high observing frequency means that pulsars - which typically exhibit steep emission spectra - are expected to be very faint. However, it also negates pulse scattering, an effect that could hinder pulsar detections in the Galactic Center. Additionally, magnetars or a secondary inverse Compton emission could be stronger at millimeter wavelengths than at lower frequencies. We present a search for pulsars close to Sgr A* using the data from the three most-sensitive stations in the EHT 2017 campaign: the Atacama Large Millimeter/submillimeter Array, the Large Millimeter Telescope and the IRAM 30 m Telescope. We apply three detection methods based on Fourier-domain analysis, the Fast-Folding-Algorithm and single pulse search targeting both pulsars and burst-like transient emission; using the simultaneity of the observations to confirm potential candidates. No new pulsars or significant bursts were found. Being the first pulsar search ever carried out at such high radio frequencies, we detail our analysis methods and give a detailed estimation of the sensitivity of the search. We conclude that the EHT 2017 observations are only sensitive to a small fraction ($\lesssim$2.2%) of the pulsars that may exist close to Sgr A*, motivating further searches for fainter pulsars in the region., Comment: 33 pages, 7 figures, 6 Tables. Accepted for publication in ApJ

Large-scale pre-trained Vision Language Models (VLMs) have proven effective for zero-shot classification. Despite the success, most traditional VLMs-based methods are restricted by the assumption of partial source supervision or ideal vocabularies, which rarely satisfy the open-world scenario. In this paper, we aim at a more challenging setting, Realistic Zero-Shot Classification, which assumes no annotation but instead a broad vocabulary. To address this challenge, we propose the Self Structural Semantic Alignment (S^3A) framework, which extracts the structural semantic information from unlabeled data while simultaneously self-learning. Our S^3A framework adopts a unique Cluster-Vote-Prompt-Realign (CVPR) algorithm, which iteratively groups unlabeled data to derive structural semantics for pseudo-supervision. Our CVPR process includes iterative clustering on images, voting within each cluster to identify initial class candidates from the vocabulary, generating discriminative prompts with large language models to discern confusing candidates, and realigning images and the vocabulary as structural semantic alignment. Finally, we propose to self-learn the CLIP image encoder with both individual and structural semantic alignment through a teacher-student learning strategy. Our comprehensive experiments across various generic and fine-grained benchmarks demonstrate that the S^3A method offers substantial improvements over existing VLMs-based approaches, achieving a more than 15% accuracy improvement over CLIP on average. Our codes, models, and prompts are publicly released at https://github.com/sheng-eatamath/S3A., Comment: AAAI'24

We propose a unified point cloud video self-supervised learning framework for object-centric and scene-centric data. Previous methods commonly conduct representation learning at the clip or frame level and cannot well capture fine-grained semantics. Instead of contrasting the representations of clips or frames, in this paper, we propose a unified self-supervised framework by conducting contrastive learning at the point level. Moreover, we introduce a new pretext task by achieving semantic alignment of superpoints, which further facilitates the representations to capture semantic cues at multiple scales. In addition, due to the high redundancy in the temporal dimension of dynamic point clouds, directly conducting contrastive learning at the point level usually leads to massive undesired negatives and insufficient modeling of positive representations. To remedy this, we propose a selection strategy to retain proper negatives and make use of high-similarity samples from other instances as positive supplements. Extensive experiments show that our method outperforms supervised counterparts on a wide range of downstream tasks and demonstrates the superior transferability of the learned representations., Comment: Accepted by ICCV 2023

Recently, the community has made tremendous progress in developing effective methods for point cloud video understanding that learn from massive amounts of labeled data. However, annotating point cloud videos is usually notoriously expensive. Moreover, training via one or only a few traditional tasks (e.g., classification) may be insufficient to learn subtle details of the spatio-temporal structure existing in point cloud videos. In this paper, we propose a Masked Spatio-Temporal Structure Prediction (MaST-Pre) method to capture the structure of point cloud videos without human annotations. MaST-Pre is based on spatio-temporal point-tube masking and consists of two self-supervised learning tasks. First, by reconstructing masked point tubes, our method is able to capture the appearance information of point cloud videos. Second, to learn motion, we propose a temporal cardinality difference prediction task that estimates the change in the number of points within a point tube. In this way, MaST-Pre is forced to model the spatial and temporal structure in point cloud videos. Extensive experiments on MSRAction-3D, NTU-RGBD, NvGesture, and SHREC'17 demonstrate the effectiveness of the proposed method., Comment: Accepted by ICCV 2023

Despite the remarkable performance of Vision Transformers (ViTs) in various visual tasks, the expanding computation and model size of ViTs have increased the demand for improved efficiency during training and inference. To address the heavy computation and parameter drawbacks, quantization is frequently studied in the community as a representative model compression technique and has seen extensive use on CNNs. However, due to the unique properties of CNNs and ViTs, the quantization applications on ViTs are still limited and underexplored. In this paper, we identify the difficulty of ViT quantization on its unique variation behaviors, which differ from traditional CNN architectures. The variations indicate the magnitude of the parameter fluctuations and can also measure outlier conditions. Moreover, the variation behaviors reflect the various sensitivities to the quantization of each module. The quantization sensitivity analysis and comparison of ViTs with CNNs help us locate the underlying differences in variations. We also find that the variations in ViTs cause training oscillations, bringing instability during quantization-aware training (QAT). Correspondingly, we solve the variation problem with an efficient knowledge-distillation-based variation-aware quantization method. The multi-crop knowledge distillation scheme can accelerate and stabilize the training and alleviate the variation's influence during QAT. We also proposed a module-dependent quantization scheme and a variation-aware regularization term to suppress the oscillation of weights. On ImageNet-1K, we obtain a 77.66% Top-1 accuracy on the extremely low-bit scenario of 2-bit Swin-T, outperforming the previous state-of-the-art quantized model by 3.35%., Comment: Code is available at https://github.com/HuangOwen/VVTQ

Observing and timing a group of millisecond pulsars (MSPs) with high rotational stability enables the direct detection of gravitational waves (GWs). The GW signals can be identified from the spatial correlations encoded in the times-of-arrival of widely spaced pulsar-pairs. The Chinese Pulsar Timing Array (CPTA) is a collaboration aiming at the direct GW detection with observations carried out using Chinese radio telescopes. This short article serves as a `table of contents' for a forthcoming series of papers related to the CPTA Data Release 1 (CPTA DR1) which uses observations from the Five-hundred-meter Aperture Spherical radio Telescope (FAST). Here, after summarizing the time span and accuracy of CPTA DR1, we report the key results of our statistical inference finding a correlated signal with amplitude $\log A_{\rm c}= -14.4 \,^{+1.0}_{-2.8}$ for spectral index in the range of $\alpha\in [-1.8, 1.5]$ assuming a GW background (GWB) induced quadrupolar correlation. The search for the Hellings-Downs (HD) correlation curve is also presented, where some evidence for the HD correlation has been found that a 4.6-$\sigma$ statistical significance is achieved using the discrete frequency method around the frequency of 14 nHz. We expect that the future International Pulsar Timing Array data analysis and the next CPTA data release will be more sensitive to the nHz GWB, which could verify the current results., Comment: 18 pages, 6 figures, submitted to "Research in astronomy and astrophysics" 22nd March 2022

We present a new dataset condensation framework termed Squeeze, Recover and Relabel (SRe$^2$L) that decouples the bilevel optimization of model and synthetic data during training, to handle varying scales of datasets, model architectures and image resolutions for efficient dataset condensation. The proposed method demonstrates flexibility across diverse dataset scales and exhibits multiple advantages in terms of arbitrary resolutions of synthesized images, low training cost and memory consumption with high-resolution synthesis, and the ability to scale up to arbitrary evaluation network architectures. Extensive experiments are conducted on Tiny-ImageNet and full ImageNet-1K datasets. Under 50 IPC, our approach achieves the highest 42.5% and 60.8% validation accuracy on Tiny-ImageNet and ImageNet-1K, outperforming all previous state-of-the-art methods by margins of 14.5% and 32.9%, respectively. Our approach also surpasses MTT in terms of speed by approximately 52$\times$ (ConvNet-4) and 16$\times$ (ResNet-18) faster with less memory consumption of 11.6$\times$ and 6.4$\times$ during data synthesis. Our code and condensed datasets of 50, 200 IPC with 4K recovery budget are available at https://github.com/VILA-Lab/SRe2L., Comment: NeurIPS 2023 spotlight. Code at https://github.com/VILA-Lab/SRe2L

We present Generalized LoRA (GLoRA), an advanced approach for universal parameter-efficient fine-tuning tasks. Enhancing Low-Rank Adaptation (LoRA), GLoRA employs a generalized prompt module to optimize pre-trained model weights and adjust intermediate activations, providing more flexibility and capability across diverse tasks and datasets. Moreover, GLoRA facilitates efficient parameter adaptation by employing a scalable, modular, layer-wise structure search that learns individual adapter of each layer. Originating from a unified mathematical formulation, GLoRA exhibits strong transfer learning, few-shot learning and domain generalization abilities, as it adapts to new tasks through not only weights but also additional dimensions like activations. Comprehensive experiments demonstrate that GLoRA outperforms all previous methods in natural, specialized, and structured vision benchmarks, achieving superior accuracy with fewer parameters and computations. The proposed method on LLaMA-1 and LLaMA-2 also show considerable enhancements compared to the original LoRA in the language domain. Furthermore, our structural re-parameterization design ensures that GLoRA incurs no extra inference cost, rendering it a practical solution for resource-limited applications. Code and models are available at: https://github.com/Arnav0400/ViT-Slim/tree/master/GLoRA., Comment: Technical report. v2: Add LLaMA-1&2 results. Code and models at https://github.com/Arnav0400/ViT-Slim/tree/master/GLoRA

We present a new self-supervised paradigm on point cloud sequence understanding. Inspired by the discriminative and generative self-supervised methods, we design two tasks, namely point cloud sequence based Contrastive Prediction and Reconstruction (CPR), to collaboratively learn more comprehensive spatiotemporal representations. Specifically, dense point cloud segments are first input into an encoder to extract embeddings. All but the last ones are then aggregated by a context-aware autoregressor to make predictions for the last target segment. Towards the goal of modeling multi-granularity structures, local and global contrastive learning are performed between predictions and targets. To further improve the generalization of representations, the predictions are also utilized to reconstruct raw point cloud sequences by a decoder, where point cloud colorization is employed to discriminate against different frames. By combining classic contrast and reconstruction paradigms, it makes the learned representations with both global discrimination and local perception. We conduct experiments on four point cloud sequence benchmarks, and report the results on action recognition and gesture recognition under multiple experimental settings. The performances are comparable with supervised methods and show powerful transferability., Comment: Accepted by AAAI2023

Self-supervised learning can extract representations of good quality from solely unlabeled data, which is appealing for point cloud videos due to their high labelling cost. In this paper, we propose a contrastive mask prediction (PointCMP) framework for self-supervised learning on point cloud videos. Specifically, our PointCMP employs a two-branch structure to achieve simultaneous learning of both local and global spatio-temporal information. On top of this two-branch structure, a mutual similarity based augmentation module is developed to synthesize hard samples at the feature level. By masking dominant tokens and erasing principal channels, we generate hard samples to facilitate learning representations with better discrimination and generalization performance. Extensive experiments show that our PointCMP achieves the state-of-the-art performance on benchmark datasets and outperforms existing full-supervised counterparts. Transfer learning results demonstrate the superiority of the learned representations across different datasets and tasks., Comment: Accepted by CVPR 2023

Interpretation of resolved polarized images of black holes by the Event Horizon Telescope (EHT) requires predictions of the polarized emission observable by an Earth-based instrument for a particular model of the black hole accretion system. Such predictions are generated by general relativistic radiative transfer (GRRT) codes, which integrate the equations of polarized radiative transfer in curved spacetime. A selection of ray-tracing GRRT codes used within the EHT collaboration is evaluated for accuracy and consistency in producing a selection of test images, demonstrating that the various methods and implementations of radiative transfer calculations are highly consistent. When imaging an analytic accretion model, we find that all codes produce images similar within a pixel-wise normalized mean squared error (NMSE) of 0.012 in the worst case. When imaging a snapshot from a cell-based magnetohydrodynamic simulation, we find all test images to be similar within NMSEs of 0.02, 0.04, 0.04, and 0.12 in Stokes I, Q, U , and V respectively. We additionally find the values of several image metrics relevant to published EHT results to be in agreement to much better precision than measurement uncertainties., Comment: Accepted for publication in ApJ

Because of its proximity and the large size of its black hole, M87 is one of the best targets for studying the launching mechanism of active galactic nucleus jets. Currently, magnetic fields are considered to be an essential factor in the launching and accelerating of the jet. However, current observational estimates of the magnetic field strength of the M87 jet are limited to the innermost part of the jet or to HST-1. No attempt has yet been made to measure the magnetic field strength in between. We aim to infer the magnetic field strength of the M87 jet out to a distance of several thousand $r_s$ by tracking the distance-dependent changes in the synchrotron spectrum of the jet from high-resolution very long baseline interferometry observations. In order to obtain high-quality spectral index maps, quasi-simultaneous observations at 22 and 43 GHz were conducted using the KVN and VERA Array (KaVA) and the VLBA. We compared the spectral index distributions obtained from the observations with a model and placed limits on the magnetic field strengths as a function of distance. The overall spectral morphology is broadly consistent over the course of these observations. The observed synchrotron spectrum rapidly steepens from $\alpha_{22-43 GHz}$ ~ -0.7 at ~ 2 mas to $\alpha_{22-43 GHz}$ ~ -2.5 at ~ 6 mas. A spectral index model in which nonthermal electron injections inside the jet decrease with distance can adequately reproduce the observed trend. This suggests the magnetic field strength of the jet at a distance of 2 - 10 mas (~ 900 $r_s$ - ~ 4500 $r_s$ in the deprojected distance) has a range of $B=(0.3 - 1.0 G)(z/2 mas)^{-0.73}$. Extrapolating to the EHT scale yields consistent results, suggesting that the majority of the magnetic flux of the jet near the black hole is preserved out to ~ 4500 $r_s$ without significant dissipation., Comment: 16 pages, 9 figures. Accepted for publication in Astronomy & Astrophysics

Introduced by Hinton et al. in 2012, dropout has stood the test of time as a regularizer for preventing overfitting in neural networks. In this study, we demonstrate that dropout can also mitigate underfitting when used at the start of training. During the early phase, we find dropout reduces the directional variance of gradients across mini-batches and helps align the mini-batch gradients with the entire dataset's gradient. This helps counteract the stochasticity of SGD and limit the influence of individual batches on model training. Our findings lead us to a solution for improving performance in underfitting models - early dropout: dropout is applied only during the initial phases of training, and turned off afterwards. Models equipped with early dropout achieve lower final training loss compared to their counterparts without dropout. Additionally, we explore a symmetric technique for regularizing overfitting models - late dropout, where dropout is not used in the early iterations and is only activated later in training. Experiments on ImageNet and various vision tasks demonstrate that our methods consistently improve generalization accuracy. Our results encourage more research on understanding regularization in deep learning and our methods can be useful tools for future neural network training, especially in the era of large data. Code is available at https://github.com/facebookresearch/dropout., Comment: ICML 2023

We report the first detection of radio recombination lines (RRLs) of ions heavier than helium. In a highly sensitive multi-band (12--50 GHz) line survey toward Orion KL with the TianMa 65-m Radio Telescope (TMRT), we successfully detected more than fifteen unblended $\alpha$ lines of RRLs of singly ionized species (XII) recombined from XIII. The Ka-band (35--50 GHz) spectrum also shows tentative signals of $\beta$ lines of ions. The detected lines can be successfully crossmatched with the the rest frequencies of RRLs of CII and/or OII. This finding greatly expands the connotation of ion RRLs, since before this work only two blended lines (105$\alpha$ and 121$\alpha$) of HeII had been reported. Our detected lines can be fitted simultaneously under assumption of local thermodynamic equilibrium (LTE). An abundance of CIII and OIII of 8.8$\times$10$^{-4}$ is obtained, avoiding the complexities of optical/infrared observations and the blending of RRLs of atoms. It is consistent with but approaches the upper bound of the value (10$^{-4}$--$10^{-3}$) estimated from optical/infrared observations. The effects of dielectronic recombination may contribute to enhancing the level populations even at large $n$. We expect future observations using radio interferometers could break the degeneracy between C and O, and help to reveal the ionization structure and dynamical evolution of various ionized regions., Comment: 3 figures, 2 tables, accepted by A&A Letter

Consistency regularization and pseudo labeling-based semi-supervised methods perform co-training using the pseudo labels from multi-view inputs. However, such co-training models tend to converge early to a consensus, degenerating to the self-training ones, and produce low-confidence pseudo labels from the perturbed inputs during training. To address these issues, we propose an Uncertainty-guided Collaborative Mean-Teacher (UCMT) for semi-supervised semantic segmentation with the high-confidence pseudo labels. Concretely, UCMT consists of two main components: 1) collaborative mean-teacher (CMT) for encouraging model disagreement and performing co-training between the sub-networks, and 2) uncertainty-guided region mix (UMIX) for manipulating the input images according to the uncertainty maps of CMT and facilitating CMT to produce high-confidence pseudo labels. Combining the strengths of UMIX with CMT, UCMT can retain model disagreement and enhance the quality of pseudo labels for the co-training segmentation. Extensive experiments on four public medical image datasets including 2D and 3D modalities demonstrate the superiority of UCMT over the state-of-the-art. Code is available at: https://github.com/Senyh/UCMT.

The number of international benchmarking competitions is steadily increasing in various fields of machine learning (ML) research and practice. So far, however, little is known about the common practice as well as bottlenecks faced by the community in tackling the research questions posed. To shed light on the status quo of algorithm development in the specific field of biomedical imaging analysis, we designed an international survey that was issued to all participants of challenges conducted in conjunction with the IEEE ISBI 2021 and MICCAI 2021 conferences (80 competitions in total). The survey covered participants' expertise and working environments, their chosen strategies, as well as algorithm characteristics. A median of 72% challenge participants took part in the survey. According to our results, knowledge exchange was the primary incentive (70%) for participation, while the reception of prize money played only a minor role (16%). While a median of 80 working hours was spent on method development, a large portion of participants stated that they did not have enough time for method development (32%). 25% perceived the infrastructure to be a bottleneck. Overall, 94% of all solutions were deep learning-based. Of these, 84% were based on standard architectures. 43% of the respondents reported that the data samples (e.g., images) were too large to be processed at once. This was most commonly addressed by patch-based training (69%), downsampling (37%), and solving 3D analysis tasks as a series of 2D tasks. K-fold cross-validation on the training set was performed by only 37% of the participants and only 50% of the participants performed ensembling based on multiple identical models (61%) or heterogeneous models (39%). 48% of the respondents applied postprocessing steps.

Although existing semi-supervised learning models achieve remarkable success in learning with unannotated in-distribution data, they mostly fail to learn on unlabeled data sampled from novel semantic classes due to their closed-set assumption. In this work, we target a pragmatic but under-explored Generalized Novel Category Discovery (GNCD) setting. The GNCD setting aims to categorize unlabeled training data coming from known and novel classes by leveraging the information of partially labeled known classes. We propose a two-stage Contrastive Affinity Learning method with auxiliary visual Prompts, dubbed PromptCAL, to address this challenging problem. Our approach discovers reliable pairwise sample affinities to learn better semantic clustering of both known and novel classes for the class token and visual prompts. First, we propose a discriminative prompt regularization loss to reinforce semantic discriminativeness of prompt-adapted pre-trained vision transformer for refined affinity relationships.Besides, we propose contrastive affinity learning to calibrate semantic representations based on our iterative semi-supervised affinity graph generation method for semantically-enhanced supervision. Extensive experimental evaluation demonstrates that our PromptCAL method is more effective in discovering novel classes even with limited annotations and surpasses the current state-of-the-art on generic and fine-grained benchmarks (e.g., with nearly 11% gain on CUB-200, and 9% on ImageNet-100) on overall accuracy. Our code is available at https://github.com/sheng-eatamath/PromptCAL., Comment: Accepted by CVPR2023

We present the study of arcsecond scale radio continuum and OH line emission of a sample of known OH megamaser galaxies with $z \geq$ 0.15 using archival Very Large Array (VLA) data. And also the results of our pilot Five hundred meter aperture spherical radio telescope (FAST) observations of 12 of these OHM galaxies. The arcsecond-scale resolution images show that the OH emission is distributed in one compact structure and spatially associated with radio continuum emission. Furthermore, nearly all the fitted components are likely smaller than the beam size ($\sim$ 1.4"), which indicates that the broad OH line profiles of these sources originated from one masing region or that more components are distributed in sub-arcsec scales. The radio parameters, including brightness temperature, spectral index, and q-index, show no significant differences with the low-redshift OHM galaxies, which have significantly lower OH line luminosities. Because these parameters are indicators of the central power sources (AGN, starburst, or both), our results indicate that the presence of radio AGN in the nuclei may not be essential for the formation of OH emission. Over 1/3 of OHMs in this sample (6/17) show possible variable features likely caused by interstellar scintillation due to small angular sizes. We might underestimate this value because these sources are associated with this sample's highest OH line flux densities. Those with low OH line flux densities might need higher sensitivity observations to study the variabilities. These results support the compact nature of OH maser emission and a starburst origin for the OHMs in our selected sample., Comment: 25 pages,7 figures,accepted by A&A