Search

Recommender systems aim to capture users' personalized preferences from the cast amount of user behaviors, making them pivotal in the era of information explosion. However, the presence of the dynamic preference, the "information cocoons", and the inherent feedback loops in recommendation make users interact with a limited number of items. Conventional recommendation algorithms typically focus on the positive historical behaviors, while neglecting the essential role of negative feedback in user interest understanding. As a promising but easy-to-ignored area, negative sampling is proficients in revealing the genuine negative aspect inherent in user behaviors, emerging as an inescapable procedure in recommendation. In this survey, we first discuss the role of negative sampling in recommendation and thoroughly analyze challenges that consistently impede its progress. Then, we conduct an extensive literature review on the existing negative sampling strategies in recommendation and classify them into five categories with their discrepant techniques. Finally, we detail the insights of the tailored negative sampling strategies in diverse recommendation scenarios and outline an overview of the prospective research directions toward which the community may engage and benefit., Comment: 38 pages, 9 figures; Under review

Image classification models often demonstrate unstable performance in real-world applications due to variations in image information, driven by differing visual perspectives of subject objects and lighting discrepancies. To mitigate these challenges, existing studies commonly incorporate additional modal information matching the visual data to regularize the model's learning process, enabling the extraction of high-quality visual features from complex image regions. Specifically, in the realm of multimodal learning, cross-modal alignment is recognized as an effective strategy, harmonizing different modal information by learning a domain-consistent latent feature space for visual and semantic features. However, this approach may face limitations due to the heterogeneity between multimodal information, such as differences in feature distribution and structure. To address this issue, we introduce a Multimodal Alignment and Reconstruction Network (MARNet), designed to enhance the model's resistance to visual noise. Importantly, MARNet includes a cross-modal diffusion reconstruction module for smoothly and stably blending information across different domains. Experiments conducted on two benchmark datasets, Vireo-Food172 and Ingredient-101, demonstrate that MARNet effectively improves the quality of image information extracted by the model. It is a plug-and-play framework that can be rapidly integrated into various image classification frameworks, boosting model performance.

The impressive capabilities in Large Language Models (LLMs) provide a powerful approach to reimagine users' typing experience. This paper demonstrates Proofread, a novel Gboard feature powered by a server-side LLM in Gboard, enabling seamless sentence-level and paragraph-level corrections with a single tap. We describe the complete system in this paper, from data generation, metrics design to model tuning and deployment. To obtain models with sufficient quality, we implement a careful data synthetic pipeline tailored to online use cases, design multifaceted metrics, employ a two-stage tuning approach to acquire the dedicated LLM for the feature: the Supervised Fine Tuning (SFT) for foundational quality, followed by the Reinforcement Learning (RL) tuning approach for targeted refinement. Specifically, we find sequential tuning on Rewrite and proofread tasks yields the best quality in SFT stage, and propose global and direct rewards in the RL tuning stage to seek further improvement. Extensive experiments on a human-labeled golden set showed our tuned PaLM2-XS model achieved 85.56\% good ratio. We launched the feature to Pixel 8 devices by serving the model on TPU v5 in Google Cloud, with thousands of daily active users. Serving latency was significantly reduced by quantization, bucket inference, text segmentation, and speculative decoding. Our demo could be seen in \href{https://youtu.be/4ZdcuiwFU7I}{Youtube}., Comment: 8 pages, 3 figures, 2 tables

Complex multi-step reasoning tasks, such as solving mathematical problems or generating code, remain a significant hurdle for even the most advanced large language models (LLMs). Verifying LLM outputs with an Outcome Reward Model (ORM) is a standard inference-time technique aimed at enhancing the reasoning performance of LLMs. However, this still proves insufficient for reasoning tasks with a lengthy or multi-hop reasoning chain, where the intermediate outcomes are neither properly rewarded nor penalized. Process supervision addresses this limitation by assigning intermediate rewards during the reasoning process. To date, the methods used to collect process supervision data have relied on either human annotation or per-step Monte Carlo estimation, both prohibitively expensive to scale, thus hindering the broad application of this technique. In response to this challenge, we propose a novel divide-and-conquer style Monte Carlo Tree Search (MCTS) algorithm named \textit{OmegaPRM} for the efficient collection of high-quality process supervision data. This algorithm swiftly identifies the first error in the Chain of Thought (CoT) with binary search and balances the positive and negative examples, thereby ensuring both efficiency and quality. As a result, we are able to collect over 1.5 million process supervision annotations to train a Process Reward Model (PRM). Utilizing this fully automated process supervision alongside the weighted self-consistency algorithm, we have enhanced the instruction tuned Gemini Pro model's math reasoning performance, achieving a 69.4\% success rate on the MATH benchmark, a 36\% relative improvement from the 51\% base model performance. Additionally, the entire process operates without any human intervention, making our method both financially and computationally cost-effective compared to existing methods., Comment: 18 pages, 5 figures, 1 table

Federated learning benefits from cross-training strategies, which enables models to train on data from distinct sources to improve the generalization capability. However, the data heterogeneity between sources may lead models to gradually forget previously acquired knowledge when undergoing cross-training to adapt to new tasks or data sources. We argue that integrating personalized and global knowledge to gather information from multiple perspectives could potentially improve performance. To achieve this goal, this paper presents a novel approach that enhances federated learning through a cross-training scheme incorporating multi-view information. Specifically, the proposed method, termed FedCT, includes three main modules, where the consistency-aware knowledge broadcasting module aims to optimize model assignment strategies, which enhances collaborative advantages between clients and achieves an efficient federated learning process. The multi-view knowledge-guided representation learning module leverages fused prototypical knowledge from both global and local views to enhance the preservation of local knowledge before and after model exchange, as well as to ensure consistency between local and global knowledge. The mixup-based feature augmentation module aggregates rich information to further increase the diversity of feature spaces, which enables the model to better discriminate complex samples. Extensive experiments were conducted on four datasets in terms of performance comparison, ablation study, in-depth analysis and case study. The results demonstrated that FedCT alleviates knowledge forgetting from both local and global views, which enables it outperform state-of-the-art methods.

Large language models (LLMs) augmented with retrieval exhibit robust performance and extensive versatility by incorporating external contexts. However, the input length grows linearly in the number of retrieved documents, causing a dramatic increase in latency. In this paper, we propose a novel paradigm named Sparse RAG, which seeks to cut computation costs through sparsity. Specifically, Sparse RAG encodes retrieved documents in parallel, which eliminates latency introduced by long-range attention of retrieved documents. Then, LLMs selectively decode the output by only attending to highly relevant caches auto-regressively, which are chosen via prompting LLMs with special control tokens. It is notable that Sparse RAG combines the assessment of each individual document and the generation of the response into a single process. The designed sparse mechanism in a RAG system can facilitate the reduction of the number of documents loaded during decoding for accelerating the inference of the RAG system. Additionally, filtering out undesirable contexts enhances the model's focus on relevant context, inherently improving its generation quality. Evaluation results of two datasets show that Sparse RAG can strike an optimal balance between generation quality and computational efficiency, demonstrating its generalizability across both short- and long-form generation tasks.

While all of the polymorphs of NH$_4$I and NH$_4$Br are non-polar, a reversible electric polarization is established in the ordered $\gamma$ phases of (NH$_4$)$_{0.73}$(ND$_4$)$_{0.27}$I and (NH$_4$)$_{0.84}$(ND$_4$)$_{0.16}$Br (where D is $^2$H) via $dc$ electric fields. The presence of two groups of orbital magnetic moments appears to be responsible for the asymmetric lattice distortions. Our findings provide an alternative pathway for hydrogen-based materials to potentially add a ferroelectric functionality., Comment: 14 pages, 3 figures

While all of the polymorphs of pure NH$_4$I and KI are non-polar, we identify that (NH$_4$)$_{0.95}$K$_{0.05}$I is ferroelectric and (NH$_4$)$_{0.87}$K$_{0.13}$I and (NH$_4$)$_{0.83}$K$_{0.17}$I are pyroelectric through measurements of their pyroelectric current and complex dielectric constant. The order to disorder phase transitions occur near 245 K. Magnetic susceptibility measurements indicate that the proton orbitals of the NH$_4$$^+$ continue to become ordered in the ground state in the (NH$_4$)$_{1-x}$K$_x$I system up to x <= 0.17. The polar phases are proposed to stem from K$^+$ ions disrupting the symmetry of proton-orbital-lattice interactions between the NH$_4$$^+$ and I$^-$ ions. Our work introduces a new pathway for the ordered phases of ammonium-based compounds to potentially become ferroelectric., Comment: 13 pages, 5 figures

The microscopic origin of the remarkable optoelectronic properties of one of the most studied contemporary materials remains unclear. Here, we identify the existence of magnetic interactions between intermolecular proton orbitals in CH$_3$NH$_3$PbI$_3$ and CH$_3$NH$_3$PbBr$_3$. In particular, a unique sharp drop and a pronounced step-up discontinuity in the magnetic susceptibility at the tetragonal-to-cubic phase transitions are identified in CH$_3$NH$_3$PbI$_3$ and CH$_3$NH$_3$PbBr$_3$, respectively. The magnetic interactions in the orthorhombic and tetragonal phases are dependent on thermal history and lattice orientation while nearly independent of the applied external magnetic field. In CH$_3$NH$_3$PbBr$_3$, the CH$_3$ and NH$_3$$^+$ components reorient in an uncorrelated fashion resulting the cubic phase to also exhibit magnetic anisotropy. Our findings provide a potential link connecting the highly light-absorbing CH$_3$NH$_3$$^+$ and the exceptional properties of the charge carriers of the inorganic framework in hybrid perovskite solar cells., Comment: Manuscript + Supplementary Material file (17 + 6 pages, 4 + 2 figures)

The $\beta$ and $\gamma$ phases of methylammonium chloride CH$_3$NH$_3$Cl and methylammonium bromide CH$_3$NH$_3$Br are identified to be ferroelectric $via$ pyroelectric current and dielectric constant measurements. The magnetic susceptibility also exhibits pronounced discontinuities at the Curie temperatures. We attribute the origin of spontaneous polarization to the emergence of two groups of proton orbital magnetic moments from the uncorrelated motion of the CH$_3$ and NH$_3$ groups in the $\beta$ and $\gamma$ phases. The two inequivalent frameworks of intermolecular orbital resonances interact with each other to distort the lattice in a non-centrosymmetric fashion. Our findings indicate that the structural instabilities in molecular frameworks are magnetic in origin as well as provide a new pathway toward uncovering new organic ferroelectrics., Comment: 5 pages, 5 figures

We observe sharp step-down discontinuities in the magnetic susceptibility of NH$_4$H$_2$PO$_4$ and NH$_4$H$_2$PO$_4$-$d$$_{60}$ (60% deuterated) along the $a$ and $c$-axes occurring exactly at their antiferroelectric transition temperatures. For the case of KH$_2$PO$_4$, less pronounced discontinuities occur at the ferroelectric transition temperature. To explain this, we treat the acid protons as individual oscillators that generate current elements which translate to magnetic forces in near resonance with each other. With decreasing temperature, the resonant forces become more commensurate which amplifies a disproportionate drop off of two types of magnetic forces to eventually trigger the structural phase transitions. For the case of NH$_4$H$_2$PO$_4$, the associated internal magnetic field appears to aid the NH$_4$$^+$ to order at higher temperature. At 49 K, a shoulder-like anomaly in both NH$_4$H$_2$PO$_4$ and KH$_2$PO$_4$ is attributed to a possible onset of macroscopic quantum tunneling of protons. Our findings bring forth a new category of intrinsic multiferroic systems., Comment: 14 pages with 5 figures

In the General Next-to-Minimal Supersymmetric Standard Model (GNMSSM), singlet particles may form a secluded sector of dark matter (DM), in which Singlino-like DM could achieve the observed relic abundance through various channels such as $\tilde{\chi}_1^0 \tilde{\chi}_1^0 \to h_s h_s, A_s A_s, h_s A_s$, where $h_s$ and $A_s$ represent singlet-dominated CP-even and CP-odd Higgs bosons. We provide analytical formulas for both the spin-independent and spin-dependent cross sections of Singlino DM scattering with nucleons, illustrating their dependence on the model's parameters in a clear manner. We also present analytic expressions for the annihilation cross sections of these three important channels. Based on these preparations, we conducted Bayesian analyses of the GNMSSM and concluded that the theory significantly favored Singlino-dominated DM over Bino-like DM across a much broader range of parameters. The combined results from our numerical analyses and the formulas distinctly highlight crucial aspects of DM physics within the GNMSSM., Comment: 44 pages, 8 tables, 6 figures

Most ferroelectric crystals have their own set of unique characteristics and ammonium sulfate (NH$_4$)$_2$SO$_4$ is no exception. We report on two previously unidentified features in ammonium sulfate: 1) that there are at least two successive transitions instead of one occurring at the Curie temperature $T$$_C$ = 223 K according to dielectric constant measurements; and 2) pronounced step-like anomalies are found in the magnetic susceptibility exactly at $T$$_C$. To explain these results, we take into account that there exists a previously unidentified linear coupling between the magnetic and electric dipole moments of the NH$_4$$^+$ tetrahedra due to their rapid reorientations and distorted geometry, respectively. The magnetic moments are small, 0.0016 $\mu$$_B$ for every $C$$_3$ reorientation which involve three protons (H$^+$) undergoing orbital motion. Nevertheless, short-range correlations exist in the paraelectric phase because the magnetic moments are restricted to only point along 14 possible orientations due to the symmetry and periodic nature of the potential wells. At $T$$_C$, $C$$_2$ reorientations (involving four protons) are no longer energetically feasible so the reduction in the degrees of freedom to 8 further enhances the effect of the magnetic interactions. This triggers long-range ordering of the orbital moments in an antiferromagnetic configuration along the $ab$-plane, which via Dzyaloshinskii-Moriya interactions, end up canting slightly toward the $c$-axis direction. Since there exists two types of inequivalent NH$_4$$^+$ groups that reorient at different frequencies with temperature and do not have the same degree of distortion, the emerging polar phase is ferrielectric., Comment: Manuscript + Supporting Information (19 + 2 pages, 6 + 2 figures)

The different types of magnetism arise mainly from how electrons move and interact with each other. In this work, we show how protons (H$^+$) also exhibit magnetic behavior. We measured the magnetic susceptibility of the ammonium halides and identified pronounced increases at 232 K, 233 K and 243 K for NH$_4$I, NH$_4$Br and NH$_4$Cl, respectively, which all coincide to the geometric ordering of its ammonium cations. With extensive literature establishing the fact that the ammonium cations exhibit rotational motion even towards the lowest temperatures, we take into account that the orbital motion of the protons carries a magnetic moment and find it to be larger than that of the paired electrons. Consequently, the structural phase transitions are magnetically-driven as the system attempts to lift 8-fold energy degeneracies of the proton orbitals via Jahn-Teller distortions. Our findings identify that NH$_4$$^+$ cations are capable of comprising magnetism which appears to be ubiquitous in ammonia-based molecular solids., Comment: Manuscript + Supporting Information file (19 + 4 pages, 5 + 3 figures). Sorry for not uploading this back in 2020!

In text-to-image generation tasks, the advancements of diffusion models have facilitated the fidelity of generated results. However, these models encounter challenges when processing text prompts containing multiple entities and attributes. The uneven distribution of attention results in the issues of entity leakage and attribute misalignment. Training from scratch to address this issue requires numerous labeled data and is resource-consuming. Motivated by this, we propose an attribution-focusing mechanism, a training-free phase-wise mechanism by modulation of attention for diffusion model. One of our core ideas is to guide the model to concentrate on the corresponding syntactic components of the prompt at distinct timesteps. To achieve this, we incorporate a temperature control mechanism within the early phases of the self-attention modules to mitigate entity leakage issues. An object-focused masking scheme and a phase-wise dynamic weight control mechanism are integrated into the cross-attention modules, enabling the model to discern the affiliation of semantic information between entities more effectively. The experimental results in various alignment scenarios demonstrate that our model attain better image-text alignment with minimal additional computational cost.

Utilizing NICER observations, we present an analysis of the soft X-ray re-brightening event of GRS 1915+105 observed in 2021. During this event, we observed the emergence of a stable, long-lasting low-frequency quasi-periodic oscillation (LFQPO) with frequencies ranging from 0.17 to 0.21 Hz. Through a careful spectral analysis, we demonstrate that a low-temperature Compton-thick gas model well characterizes the emitted radiation. By examining the spectrum and identifying numerous absorption lines, we discerned a transition in the wind properties. This transition was marked by a shift from a state characterized by low speed, high column density, and high ionization degree to one featuring still low speed but low column density and ionization degree. Intriguingly, the presence or absence of the QPO signal is perfectly correlated with these distinct wind characteristics. The low-speed wind observed could be indicative of a 'failed wind', while the observed shift implies a transition from a magnetically to a thermally driven wind. Notably, this QPO signal exclusively manifested itself during the magnetically driven phase, suggesting the possibility of a novel perturbation associated with magnetic effects.

The development of virtual try-on has revolutionized online shopping by allowing customers to visualize themselves in various fashion items, thus extending the in-store try-on experience to the cyber space. Although virtual try-on has attracted considerable research initiatives, existing systems only focus on the quality of image generation, overlooking whether the fashion item is a good match to the given person and clothes. Recognizing this gap, we propose to design a one-stop Smart Fitting Room, with the novel formulation of matching-aware virtual try-on. Following this formulation, we design a Hybrid Matching-aware Virtual Try-On Framework (HMaVTON), which combines retrieval-based and generative methods to foster a more personalized virtual try-on experience. This framework integrates a hybrid mix-and-match module and an enhanced virtual try-on module. The former can recommend fashion items available on the platform to boost sales and generate clothes that meets the diverse tastes of consumers. The latter provides high-quality try-on effects, delivering a one-stop shopping service. To validate the effectiveness of our approach, we enlist the expertise of fashion designers for a professional evaluation, assessing the rationality and diversity of the clothes combinations and conducting an evaluation matrix analysis. Our method significantly enhances the practicality of virtual try-on. The code is available at https://github.com/Yzcreator/HMaVTON.

Reinforcement learning (RL) can align language models with non-differentiable reward signals, such as human preferences. However, a major challenge arises from the sparsity of these reward signals - typically, there is only a single reward for an entire output. This sparsity of rewards can lead to inefficient and unstable learning. To address this challenge, our paper introduces an novel framework that utilizes the critique capability of Large Language Models (LLMs) to produce intermediate-step rewards during RL training. Our method involves coupling a policy model with a critic language model, which is responsible for providing comprehensive feedback of each part of the output. This feedback is then translated into token or span-level rewards that can be used to guide the RL training process. We investigate this approach under two different settings: one where the policy model is smaller and is paired with a more powerful critic model, and another where a single language model fulfills both roles. We assess our approach on three text generation tasks: sentiment control, language model detoxification, and summarization. Experimental results show that incorporating artificial intrinsic rewards significantly improve both sample efficiency and the overall performance of the policy model, supported by both automatic and human evaluation.

Pioneering efforts have verified the effectiveness of the diffusion models in exploring the informative uncertainty for recommendation. Considering the difference between recommendation and image synthesis tasks, existing methods have undertaken tailored refinements to the diffusion and reverse process. However, these approaches typically use the highest-score item in corpus for user interest prediction, leading to the ignorance of the user's generalized preference contained within other items, thereby remaining constrained by the data sparsity issue. To address this issue, this paper presents a novel Plug-in Diffusion Model for Recommendation (PDRec) framework, which employs the diffusion model as a flexible plugin to jointly take full advantage of the diffusion-generating user preferences on all items. Specifically, PDRec first infers the users' dynamic preferences on all items via a time-interval diffusion model and proposes a Historical Behavior Reweighting (HBR) mechanism to identify the high-quality behaviors and suppress noisy behaviors. In addition to the observed items, PDRec proposes a Diffusion-based Positive Augmentation (DPA) strategy to leverage the top-ranked unobserved items as the potential positive samples, bringing in informative and diverse soft signals to alleviate data sparsity. To alleviate the false negative sampling issue, PDRec employs Noise-free Negative Sampling (NNS) to select stable negative samples for ensuring effective model optimization. Extensive experiments and analyses on four datasets have verified the superiority of the proposed PDRec over the state-of-the-art baselines and showcased the universality of PDRec as a flexible plugin for commonly-used sequential encoders in different recommendation scenarios. The code is available in https://github.com/hulkima/PDRec., Comment: Accepted by AAAI 2024

Evaluating natural language systems poses significant challenges, particularly in the realms of natural language understanding and high-level reasoning. In this paper, we introduce 'Fusion-Eval', an innovative approach that leverages Large Language Models (LLMs) to integrate insights from various assistant evaluators. The LLM is given the example to evaluate along with scores from the assistant evaluators. Each of these evaluators specializes in assessing distinct aspects of responses. Fusion-Eval achieves a 0.962 system-level Kendall-Tau correlation with humans on SummEval and a 0.744 turn-level Spearman correlation on TopicalChat, which is significantly higher than baseline methods. These results highlight Fusion-Eval's significant potential in the realm of natural language system evaluation.

Parameter Efficient Tuning has been an prominent approach to adapt the Large Language Model to downstream tasks. Most previous works considers adding the dense trainable parameters, where all parameters are used to adapt certain task. We found this less effective empirically using the example of LoRA that introducing more trainable parameters does not help. Motivated by this we investigate the importance of leveraging "sparse" computation and propose SiRA: sparse mixture of low rank adaption. SiRA leverages the Sparse Mixture of Expert(SMoE) to boost the performance of LoRA. Specifically it enforces the top $k$ experts routing with a capacity limit restricting the maximum number of tokens each expert can process. We propose a novel and simple expert dropout on top of gating network to reduce the over-fitting issue. Through extensive experiments, we verify SiRA performs better than LoRA and other mixture of expert approaches across different single tasks and multitask settings.

The CMS and ATLAS collaborations recently published their results searching for light Higgs bosons, using the complete Run 2 data of the LHC. Both reported an excess in the diphoton invariant mass distribution at $m_{\gamma \gamma} \simeq 95.4~{\rm GeV}$ with compatible signal strengths. The combined result corresponded to a local significance of $3.1\sigma$. Besides, the mass of the diphoton signal coincided with that of the $b\bar{b}$ excess observed at the LEP. Given the remarkable theoretical advantages of the general Next-to-Minimal Supersymmetric Standard Model, we interpret these excesses by the resonant productions of the singlet-dominated CP-even Higgs boson predicted by the theory. Using both analytic formulae and numerical results, we show that the idea can interpret the excesses by broad parameter space without contradicting current experimental restrictions, including those from the 125~{\rm GeV} Higgs data, the dark matter relic abundance and direct detection experiments, and the collider searches for supersymmetry and extra Higgs bosons. Although the explanations are scarcely affected by present Higgs data and the LHC search for supersymmetry, the dark matter physics may leave footprints on them. We also survey the other signals of the light Higgs boson at the LHC., Comment: 34 pages, 11 figures

Critical thinking is essential for rational decision-making and problem-solving. This skill hinges on the ability to provide precise and reasoned critiques and is a hallmark of human intelligence. In the era of large language models (LLMs), this study explores the ability of LLMs to deliver accurate critiques across various tasks. We are interested in this topic as a capable critic model could not only serve as a reliable evaluator, but also as a source of supervised signals for model tuning. Particularly, if a model can self-critique, it has the potential for autonomous self-improvement. To examine this, we introduce a unified evaluation framework for assessing the critique abilities of LLMs. We develop a benchmark called CriticBench, which comprises 3K high-quality natural language queries and corresponding model responses; and annotate the correctness of these responses. The benchmark cover tasks such as math problem-solving, code completion, and question answering. We evaluate multiple LLMs on the collected dataset and our analysis reveals several noteworthy insights: (1) Critique is generally challenging for most LLMs, and this capability often emerges only when models are sufficiently large. (2) In particular, self-critique is especially difficult. Even top-performing LLMs struggle to achieve satisfactory performance. (3) Models tend to have lower critique accuracy on problems where they are most uncertain. To this end, we introduce a simple yet effective baseline named self-check, which leverages self-critique to improve task performance for various models. We hope this study serves as an initial exploration into understanding the critique abilities of LLMs, and aims to inform future research, including the development of more proficient critic models and the application of critiques across diverse tasks.

Large Language Models (LLMs) have demonstrated impressive capabilities for text rewriting. Nonetheless, the large sizes of these models make them impractical for on-device inference, which would otherwise allow for enhanced privacy and economical inference. Creating a smaller yet potent language model for text rewriting presents a formidable challenge because it requires balancing the need for a small size with the need to retain the emergent capabilities of the LLM, that requires costly data collection. To address the above challenge, we introduce a new instruction tuning approach for building a mobile-centric text rewriting model. Our strategies enable the generation of high quality training data without any human labeling. In addition, we propose a heuristic reinforcement learning framework which substantially enhances performance without requiring preference data. To further bridge the performance gap with the larger server-side model, we propose an effective approach that combines the mobile rewrite agent with the server model using a cascade. To tailor the text rewriting tasks to mobile scenarios, we introduce MessageRewriteEval, a benchmark that focuses on text rewriting for messages through natural language instructions. Through empirical experiments, we demonstrate that our on-device model surpasses the current state-of-the-art LLMs in text rewriting while maintaining a significantly reduced model size. Notably, we show that our proposed cascading approach improves model performance.

Representation learning for images has been advanced by recent progress in more complex neural models such as the Vision Transformers and new learning theories such as the structural causal models. However, these models mainly rely on the classification loss to implicitly regularize the class-level data distributions, and they may face difficulties when handling classes with diverse visual patterns. We argue that the incorporation of the structural information between data samples may improve this situation. To achieve this goal, this paper presents a framework termed \textbf{C}lass-level Structural Relation Modeling and Smoothing for Visual Representation Learning (CSRMS), which includes the Class-level Relation Modelling, Class-aware Graph Sampling, and Relational Graph-Guided Representation Learning modules to model a relational graph of the entire dataset and perform class-aware smoothing and regularization operations to alleviate the issue of intra-class visual diversity and inter-class similarity. Specifically, the Class-level Relation Modelling module uses a clustering algorithm to learn the data distributions in the feature space and identify three types of class-level sample relations for the training set; Class-aware Graph Sampling module extends typical training batch construction process with three strategies to sample dataset-level sub-graphs; and Relational Graph-Guided Representation Learning module employs a graph convolution network with knowledge-guided smoothing operations to ease the projection from different visual patterns to the same class. Experiments demonstrate the effectiveness of structured knowledge modelling for enhanced representation learning and show that CSRMS can be incorporated with any state-of-the-art visual representation learning models for performance gains. The source codes and demos have been released at https://github.com/czt117/CSRMS.

Federated Learning aims to learn a global model on the server side that generalizes to all clients in a privacy-preserving manner, by leveraging the local models from different clients. Existing solutions focus on either regularizing the objective functions among clients or improving the aggregation mechanism for the improved model generalization capability. However, their performance is typically limited by the dataset biases, such as the heterogeneous data distributions and the missing classes. To address this issue, this paper presents a cross-silo prototypical calibration method (FedCSPC), which takes additional prototype information from the clients to learn a unified feature space on the server side. Specifically, FedCSPC first employs the Data Prototypical Modeling (DPM) module to learn data patterns via clustering to aid calibration. Subsequently, the cross-silo prototypical calibration (CSPC) module develops an augmented contrastive learning method to improve the robustness of the calibration, which can effectively project cross-source features into a consistent space while maintaining clear decision boundaries. Moreover, the CSPC module's ease of implementation and plug-and-play characteristics make it even more remarkable. Experiments were conducted on four datasets in terms of performance comparison, ablation study, in-depth analysis and case study, and the results verified that FedCSPC is capable of learning the consistent features across different data sources of the same class under the guidance of calibrated model, which leads to better performance than the state-of-the-art methods. The source codes have been released at https://github.com/qizhuang-qz/FedCSPC.

Inspired by the recent measurements of the muon and electron anomalous magnetic moments, the rapid progress of the LHC search for supersymmetry, and the significantly improved sensitivities of dark matter direct detection experiments, we studied the supersymmetric contribution to the electron \texorpdfstring{$g-2$}{}, $a_e^{\rm SUSY}$, in the Next-to-Minimal Supersymmetric Standard Model with a discrete $\mathbb{Z}_3$ symmetry. We concluded that $a_e^{\rm SUSY}$ was mainly correlated with $a_\mu^{\rm SUSY}$ by the formula $a_e^{\rm SUSY}/m_e^2 \simeq a_\mu^{\rm SUSY}/m_\mu^2$, and significant violations of this correlation might occur only in rare cases. As a result, $a_e^{\rm SUSY}$ was typically around $5 \times 10^{-14}$ when $a_\mu^{\rm SUSY} \simeq 2.5 \times 10^{-9}$. We also concluded that the dark matter direct detection and LHC experiments played crucial roles in determining the maximum reach of $a_e^{\rm SUSY}$. Concretely, $a_e^{\rm SUSY}$ might be around $3 \times 10^{-13}$ in the optimum cases if one used the XENON-1T experiment to limit the supersymmetry parameter space. This prediction, however, was reduced to $1.5 \times 10^{-13}$ after implementing the LZ restrictions and $1.0 \times 10^{-13}$ when further considering the LHC restrictions., Comment: 38 pages, 4 figures, 6 tables

Large Language Models (LLMs) have demonstrated impressive capabilities in creative tasks such as storytelling and E-mail generation. However, as LLMs are primarily trained on final text results rather than intermediate revisions, it might be challenging for them to perform text rewriting tasks. Most studies in the rewriting tasks focus on a particular transformation type within the boundaries of single sentences. In this work, we develop new strategies for instruction tuning and reinforcement learning to better align LLMs for cross-sentence rewriting tasks using diverse wording and structures expressed through natural languages including 1) generating rewriting instruction data from Wiki edits and public corpus through instruction generation and chain-of-thought prompting; 2) collecting comparison data for reward model training through a new ranking function. To facilitate this research, we introduce OpenRewriteEval, a novel benchmark covers a wide variety of rewriting types expressed through natural language instructions. Our results show significant improvements over a variety of baselines. The public repository is available on GitHub under Google Research (https://github.com/google-research/google-research/tree/master/rewritelm).

Cross-domain recommendation (CDR) aims to leverage the correlation of users' behaviors in both the source and target domains to improve the user preference modeling in the target domain. Conventional CDR methods typically explore the dual-relations between the source and target domains' behaviors. However, this may ignore the informative mixed behaviors that naturally reflect the user's global preference. To address this issue, we present a novel framework, termed triple sequence learning for cross-domain recommendation (Tri-CDR), which jointly models the source, target, and mixed behavior sequences to highlight the global and target preference and precisely model the triple correlation in CDR. Specifically, Tri-CDR independently models the hidden representations for the triple behavior sequences and proposes a triple cross-domain attention (TCA) method to emphasize the informative knowledge related to both user's global and target-domain preference. To comprehensively explore the cross-domain correlations, we design a triple contrastive learning (TCL) strategy that simultaneously considers the coarse-grained similarities and fine-grained distinctions among the triple sequences, ensuring the alignment while preserving information diversity in multi-domain. We conduct extensive experiments and analyses on six cross-domain settings. The significant improvements of Tri-CDR with different sequential encoders verify its effectiveness and universality. The code will be released upon acceptance., Comment: 24 pages, 6 figures, under review

Motivated by the recent measurement of muon anomalous magnetic moment at Fermilab, the rapid progress of the LHC search for supersymmetry, and the significantly improved sensitivities of dark matter direct detection experiments, we studied their impacts on the Minimal Supersymmetric Standard Model (MSSM). We conclude that higgsino mass should be larger than about $500~{\rm GeV}$ for $M_1 < 0 $ and $630~{\rm GeV}$ for $M_1 > 100~{\rm GeV}$, where $M_1$ denotes the bino mass. These improved bounds imply a tuning of ${\cal{O}}(1\%)$ to predict the $Z$-boson mass and simultaneously worsen the naturalness of the $Z$- and $h$-mediated resonant annihilations to achieve the measured dark matter density. We also conclude that the LHC restrictions have set lower bounds on the sparticle mass spectra: $ m_{\tilde{\chi}_1^0} \gtrsim 210~{\rm GeV}$, $m_{\tilde{\chi}_2^0}, m_{\tilde{\chi}_1^\pm} \gtrsim 235~{\rm GeV}$, $m_{\tilde{\chi}_3^0} \gtrsim 515~{\rm GeV}$, $m_{\tilde{\chi}_4^0} \gtrsim 525~{\rm GeV}$, $m_{\tilde{\chi}_2^\pm} \gtrsim 530~{\rm GeV}$, $m_{\tilde{\nu}_\mu} \gtrsim 235~{\rm GeV}$, $ m_{\tilde{\mu}_1} \gtrsim 215~{\rm GeV}$, and $m_{\tilde{\mu}_2} \gtrsim 250~{\rm GeV}$, where $\tilde{\chi}_{2}^0$ and $\tilde{\chi}_1^\pm$ are wino-dominated when they are lighter than about $500~{\rm GeV}$. These bounds are far beyond the reach of the LEP experiments in searching for supersymmetry and have not been acquired before. In addition, we illuminate how some parameter spaces of the MSSM have been tested at the LHC and provide five scenarios in which the theory coincides with the LHC restrictions. Once the muon g-2 anomaly is confirmed to originate from supersymmetry, this research may serve as a guide to explore the characteristics of the MSSM in future experiments., Comment: 41 pages, 7 figures, 10 tables

With the rapid progress of dark matter direct detection experiments, the attractiveness of the popular bino-dominated dark matter in economical supersymmetric theories is fading. As an alternative, the singlino-dominated dark matter in general Next-to-Minimal Supersymmetric Standard Model (NMSSM) is paying due attention. This scenario has the following distinct characteristics: free from the tadpole problem and the domain-wall problem of the NMSSM with a $Z_3$-symmetry, predicting more stable vacuum states than the $Z_3$-NMSSM, capable of forming an economical secluded dark matter sector to yield the dark matter experimental results naturally, and readily weaken the restrictions from the LHC search for SUSY. Consequently, it can explain the muon g-2 anomaly in broad parameter space that agrees with various experimental results while simultaneously breaking the electroweak symmetry naturally. In this study, we show in detail how the scenario coincides with the experiments, such as the SUSY search at the LHC, the dark matter search by the LZ experiment, and the improved measurement of the muon g-2. We provide a simple and clear picture of the physics inherent in the general NMSSM.

Chapter generation becomes practical technique for online videos nowadays. The chapter breakpoints enable users to quickly find the parts they want and get the summative annotations. However, there is no public method and dataset for this task. To facilitate the research along this direction, we introduce a new dataset called Chapter-Gen, which consists of approximately 10k user-generated videos with annotated chapter information. Our data collection procedure is fast, scalable and does not require any additional manual annotation. On top of this dataset, we design an effective baseline specificlly for video chapters generation task. which captures two aspects of a video,including visual dynamics and narration text. It disentangles local and global video features for localization and title generation respectively. To parse the long video efficiently, a skip sliding window mechanism is designed to localize potential chapters. And a cross attention multi-modal fusion module is developed to aggregate local features for title generation. Our experiments demonstrate that the proposed framework achieves superior results over existing methods which illustrate that the method design for similar task cannot be transfered directly even after fine-tuning. Code and dataset are available at https://github.com/czt117/MVCG.

The new measurement of $W$-boson mass by the CDF collaboration revealed a remarkable $7\sigma$ disagreement with the Standard Model (SM) prediction. If confirmed by other experiments, then the disagreement strongly indicates the existence of new physics beyond the SM. In this work, seven vectorlike quark (VLQ) extensions of the SM are investigated to interpret the anomaly, and it is found that three can explain the anomaly in broad parameter space. The explanations are consistent with the constraints from oblique parameters, the LHC search for VLQs, the measurements of the properties for the top quark, bottom quark, and Higgs boson, and the perturbativity criterion. The typical size of the involved Yukawa coupling is around 1, which is comparable to the top quark Yukawa coupling in the SM. The other extensions, however, either predict a negative correction to the mass in reasonable parameter space or explain the anomaly by unnatural theoretical input parameters., Comment: This article should be an updated version of 2204.09477