Search

Significant advancements have been made by Large Language Models (LLMs) in the domains of natural language understanding and automated content creation. However, they still face persistent problems, including substantial computational costs and inadequate availability of training data. The combination of Federated Learning (FL) and LLMs (federated LLMs) offers a solution by leveraging distributed data while protecting privacy, which positions it as an ideal choice for sensitive domains. However, Federated LLMs still suffer from robustness challenges, including data heterogeneity, malicious clients, and adversarial attacks, which greatly hinder their applications. We first introduce the robustness problems in federated LLMs, to address these challenges, we propose FedEAT (Federated Embedding space Adversarial Training), a novel framework that applies adversarial training in the embedding space of client LLM and employs a robust aggregation approach, specifically geometric median aggregation, to enhance the robustness of Federated LLMs. Our experiments demonstrate that FedEAT effectively improves the robustness of Federated LLMs with minimal performance loss.

Operations Research (OR) is vital for decision-making in many industries. While recent OR methods have seen significant improvements in automation and efficiency through integrating Large Language Models (LLMs), they still struggle to produce meaningful explanations. This lack of clarity raises concerns about transparency and trustworthiness in OR applications. To address these challenges, we propose a comprehensive framework, Explainable Operations Research (EOR), emphasizing actionable and understandable explanations accompanying optimization. The core of EOR is the concept of Decision Information, which emerges from what-if analysis and focuses on evaluating the impact of complex constraints (or parameters) changes on decision-making. Specifically, we utilize bipartite graphs to quantify the changes in the OR model and adopt LLMs to improve the explanation capabilities. Additionally, we introduce the first industrial benchmark to rigorously evaluate the effectiveness of explanations and analyses in OR, establishing a new standard for transparency and clarity in the field.

Federated Foundation Models (FedFMs) represent a distributed learning paradigm that fuses general competences of foundation models as well as privacy-preserving capabilities of federated learning. This combination allows the large foundation models and the small local domain models at the remote clients to learn from each other in a teacher-student learning setting. This paper provides a comprehensive summary of the ten challenging problems inherent in FedFMs, encompassing foundational theory, utilization of private data, continual learning, unlearning, Non-IID and graph data, bidirectional knowledge transfer, incentive mechanism design, game mechanism design, model watermarking, and efficiency. The ten challenging problems manifest in five pivotal aspects: ``Foundational Theory," which aims to establish a coherent and unifying theoretical framework for FedFMs. ``Data," addressing the difficulties in leveraging domain-specific knowledge from private data while maintaining privacy; ``Heterogeneity," examining variations in data, model, and computational resources across clients; ``Security and Privacy," focusing on defenses against malicious attacks and model theft; and ``Efficiency," highlighting the need for improvements in training, communication, and parameter efficiency. For each problem, we offer a clear mathematical definition on the objective function, analyze existing methods, and discuss the key challenges and potential solutions. This in-depth exploration aims to advance the theoretical foundations of FedFMs, guide practical implementations, and inspire future research to overcome these obstacles, thereby enabling the robust, efficient, and privacy-preserving FedFMs in various real-world applications.

This paper addresses generalized category discovery (GCD), the task of clustering unlabeled data from potentially known or unknown categories with the help of labeled instances from each known category. Compared to traditional semi-supervised learning, GCD is more challenging because unlabeled data could be from novel categories not appearing in labeled data. Current state-of-the-art methods typically learn a parametric classifier assisted by self-distillation. While being effective, these methods do not make use of cross-instance similarity to discover class-specific semantics which are essential for representation learning and category discovery. In this paper, we revisit the association-based paradigm and propose a Prior-constrained Association Learning method to capture and learn the semantic relations within data. In particular, the labeled data from known categories provides a unique prior for the association of unlabeled data. Unlike previous methods that only adopts the prior as a pre or post-clustering refinement, we fully incorporate the prior into the association process, and let it constrain the association towards a reliable grouping outcome. The estimated semantic groups are utilized through non-parametric prototypical contrast to enhance the representation learning. A further combination of both parametric and non-parametric classification complements each other and leads to a model that outperforms existing methods by a significant margin. On multiple GCD benchmarks, we perform extensive experiments and validate the effectiveness of our proposed method., Comment: Accepted to AAAI 2025

Domain-generalizable re-identification (DG Re-ID) aims to train a model on one or more source domains and evaluate its performance on unseen target domains, a task that has attracted growing attention due to its practical relevance. While numerous methods have been proposed, most rely on discriminative or contrastive learning frameworks to learn generalizable feature representations. However, these approaches often fail to mitigate shortcut learning, leading to suboptimal performance. In this work, we propose a novel method called diffusion model-assisted representation learning with a correlation-aware conditioning scheme (DCAC) to enhance DG Re-ID. Our method integrates a discriminative and contrastive Re-ID model with a pre-trained diffusion model through a correlation-aware conditioning scheme. By incorporating ID classification probabilities generated from the Re-ID model with a set of learnable ID-wise prompts, the conditioning scheme injects dark knowledge that captures ID correlations to guide the diffusion process. Simultaneously, feedback from the diffusion model is back-propagated through the conditioning scheme to the Re-ID model, effectively improving the generalization capability of Re-ID features. Extensive experiments on both single-source and multi-source DG Re-ID tasks demonstrate that our method achieves state-of-the-art performance. Comprehensive ablation studies further validate the effectiveness of the proposed approach, providing insights into its robustness. Codes will be available at https://github.com/RikoLi/DCAC.

When multiple influencers attempt to compete for a receiver's attention, their influencing strategies must account for the presence of one another. We introduce the Battling Influencers Game (BIG), a multi-player simultaneous-move general-sum game, to provide a game-theoretic characterization of this social phenomenon. We prove that BIG is a potential game, that it has either one or an infinite number of pure Nash equilibria (NEs), and these pure NEs can be found by convex optimization. Interestingly, we also prove that at any pure NE, all (except at most one) influencers must exaggerate their actions to the maximum extent. In other words, it is rational for the influencers to be non-truthful and extreme because they anticipate other influencers to cancel out part of their influence. We discuss the implications of BIG to value alignment., Comment: 9 pages, 8 figures

We introduce the concept of parabolic bases to establish a localized framework for parabolic bundles and parabolic $\lambda$-connections. Building on this foundation, we propose a novel method for constructing the parabolic non-abelian Hodge correspondence in positive characteristic, extending the work originally developed by Krishnamoorthy and Sheng for algebraic curves. Additionally, we investigate the rank $2$ parabolic Higgs-de Rham flow operator and present a modified version of the Sun-Yang-Zuo algorithm, specifically adapted to the parabolic setting., Comment: The paper has been accepted for publication in SCIENCE CHINA Mathematics

We introduce our first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1. DeepSeek-R1-Zero, a model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step, demonstrates remarkable reasoning capabilities. Through RL, DeepSeek-R1-Zero naturally emerges with numerous powerful and intriguing reasoning behaviors. However, it encounters challenges such as poor readability, and language mixing. To address these issues and further enhance reasoning performance, we introduce DeepSeek-R1, which incorporates multi-stage training and cold-start data before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1-1217 on reasoning tasks. To support the research community, we open-source DeepSeek-R1-Zero, DeepSeek-R1, and six dense models (1.5B, 7B, 8B, 14B, 32B, 70B) distilled from DeepSeek-R1 based on Qwen and Llama.

The application of large language models (LLMs) has achieved remarkable success in various fields, but their effectiveness in specialized domains like the Chinese insurance industry remains underexplored. The complexity of insurance knowledge, encompassing specialized terminology and diverse data types, poses significant challenges for both models and users. To address this, we introduce InsQABench, a benchmark dataset for the Chinese insurance sector, structured into three categories: Insurance Commonsense Knowledge, Insurance Structured Database, and Insurance Unstructured Documents, reflecting real-world insurance question-answering tasks.We also propose two methods, SQL-ReAct and RAG-ReAct, to tackle challenges in structured and unstructured data tasks. Evaluations show that while LLMs struggle with domain-specific terminology and nuanced clause texts, fine-tuning on InsQABench significantly improves performance. Our benchmark establishes a solid foundation for advancing LLM applications in the insurance domain, with data and code available at https://github.com/HaileyFamo/InsQABench.git.

We characterize $\tau$-tilting modules as $1$-tilting modules over quotient algebras satisfying a tensor-vanishing condition, and characterize $1$-tilting modules as $\tau$-tilting modules satisfying a ${\rm Tor}^1$-vanishing condition. We use delooping levels to study \emph{Self-orthogonal $\tau$-tilting Conjecture}: any self-orthogonal $\tau$-tilting module is $1$-tilting. We confirm the conjecture when the endomorphism algebra of the module has finite global delooping level., Comment: 10 Pages. Any comments are welcome

With the rise of multimodal large language models, accurately extracting and understanding textual information from video content, referred to as video based optical character recognition (Video OCR), has become a crucial capability. This paper introduces a novel benchmark designed to evaluate the video OCR performance of multi-modal models in videos. Comprising 1,028 videos and 2,961 question-answer pairs, this benchmark proposes several key challenges through 6 distinct subtasks: (1) Recognition of text content itself and its basic visual attributes, (2)Semantic and Spatial Comprehension of OCR objects in videos (3) Dynamic Motion detection and Temporal Localization. We developed this benchmark using a semi-automated approach that integrates the OCR ability of image LLMs with manual refinement, balancing efficiency, cost, and data quality. Our resource aims to help advance research in video LLMs and underscores the need for improving OCR ability for video LLMs. The benchmark will be released on https://github.com/YuHuiGao/FG-Bench.git., Comment: Accepted by CoLing 2025 (The 31st International Conference on Computational Linguistics)

With the advancement of aerospace technology and the increasing demands of military applications, the development of low false-alarm and high-precision infrared small target detection algorithms has emerged as a key focus of research globally. However, the traditional model-driven method is not robust enough when dealing with features such as noise, target size, and contrast. The existing deep-learning methods have limited ability to extract and fuse key features, and it is difficult to achieve high-precision detection in complex backgrounds and when target features are not obvious. To solve these problems, this paper proposes a deep-learning infrared small target detection method that combines image super-resolution technology with multi-scale observation. First, the input infrared images are preprocessed with super-resolution and multiple data enhancements are performed. Secondly, based on the YOLOv5 model, we proposed a new deep-learning network named YOLO-MST. This network includes replacing the SPPF module with the self-designed MSFA module in the backbone, optimizing the neck, and finally adding a multi-scale dynamic detection head to the prediction head. By dynamically fusing features from different scales, the detection head can better adapt to complex scenes. The mAP@0.5 detection rates of this method on two public datasets, SIRST and IRIS, reached 96.4% and 99.5% respectively, more effectively solving the problems of missed detection, false alarms, and low precision.

We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were thoroughly validated in DeepSeek-V2. Furthermore, DeepSeek-V3 pioneers an auxiliary-loss-free strategy for load balancing and sets a multi-token prediction training objective for stronger performance. We pre-train DeepSeek-V3 on 14.8 trillion diverse and high-quality tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages to fully harness its capabilities. Comprehensive evaluations reveal that DeepSeek-V3 outperforms other open-source models and achieves performance comparable to leading closed-source models. Despite its excellent performance, DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training. In addition, its training process is remarkably stable. Throughout the entire training process, we did not experience any irrecoverable loss spikes or perform any rollbacks. The model checkpoints are available at https://github.com/deepseek-ai/DeepSeek-V3.

Gradient leakage attacks pose a significant threat to the privacy guarantees of federated learning. While distortion-based protection mechanisms are commonly employed to mitigate this issue, they often lead to notable performance degradation. Existing methods struggle to preserve model performance while ensuring privacy. To address this challenge, we propose a novel data augmentation-based framework designed to achieve a favorable privacy-utility trade-off, with the potential to enhance model performance in certain cases. Our framework incorporates the AugMix algorithm at the client level, enabling data augmentation with controllable severity. By integrating the Jensen-Shannon divergence into the loss function, we embed the distortion introduced by AugMix into the model gradients, effectively safeguarding privacy against deep leakage attacks. Moreover, the JS divergence promotes model consistency across different augmentations of the same image, enhancing both robustness and performance. Extensive experiments on benchmark datasets demonstrate the effectiveness and stability of our method in protecting privacy. Furthermore, our approach maintains, and in some cases improves, model performance, showcasing its ability to achieve a robust privacy-utility trade-off.

We study an inflation model with nonminimal derivative coupling that features a coupling between the derivative of the inflaton field and the Einstein tensor. This model naturally amplifies curvature perturbations at small scales via gravitationally enhanced friction, a mechanism critical for the formation of primordial black holes and the associated production of potentially detectable scalar-induced gravitational waves. We derive analytical expressions for the primordial power spectrum, enabling efficient exploration of the model parameter space without requiring computationally intensive numerical solutions of the Mukhanov-Sasaki equation. Using the third data release of the Parkes Pulsar Timing Array (PPTA DR3), we constrain the model parameters characterizing the coupling function: $\phi_c = 3.7^{+0.3}_{-0.5} M_\mathrm{P}$, $\log_{10} \omega_L = 7.1^{+0.6}_{-0.3}$, and $\log_{10} \sigma = -8.3^{+0.3}_{-0.6}$ at 90\% confidence level. Our results demonstrate the growing capability of pulsar timing arrays to probe early Universe physics, complementing traditional cosmic microwave background observations by providing unique constraints on inflationary dynamics at small scales., Comment: 14 pages, 5figures. Accepted for publication as a Letter in Physical Review D

The birth mass function of neutron stars encodes rich information about supernova explosions, double star evolution, and properties of matter under extreme conditions. To date, it has remained poorly constrained by observations, however. Applying probabilistic corrections to account for mass accreted by recycled pulsars in binary systems to mass measurements of 90 neutron stars, we find that the birth masses of neutron stars can be described by a unimodal distribution that smoothly turns on at $\mathbf{\unit[1.1]{\mathrm{M}_{\odot}}}$, peaks at $\mathbf{\approx \unit[1.27]{\mathrm{M}_{\odot}}}$, before declining as a steep power law. Such a ``turn-on" power-law distribution is strongly favoured against the widely-adopted empirical double-Gaussian model at the $\mathbf{3\sigma}$ level. The power-law shape may be inherited from the initial mass function of massive stars, but the relative dearth of massive neutron stars implies that single stars with initial masses greater than $\mathbf{\approx \unit[18]{\mathrm{M}_{\odot}}}$ do not form neutron stars, in agreement with the absence of massive red supergiant progenitors to supernovae., Comment: 57 pages, 14 figures, 4 tables

While traditional game models often simplify interactions among agents as static, real-world social relationships are inherently dynamic, influenced by both immediate payoffs and alternative information. Motivated by this fact, we introduce a coevolutionary multiplex network model that incorporates the concepts of a relationship threshold and a recommendation mechanism to explore how the strength of relationships among agents interacts with their strategy choices within the framework of weak prisoner's dilemma games. In the relationship layer, the relationship strength between agents varies based on interaction outcomes. In return, the strategy choice of agents in the game layer is influenced by both payoffs and relationship indices, and agents can interact with distant agents through a recommendation mechanism. Simulation of various network topologies reveals that a higher average degree supports cooperation, although increased randomness in interactions may inhibit its formation. Interestingly, a higher threshold value of interaction quality is detrimental, while the applied recommendation protocol can improve global cooperation. The best results are obtained when the relative weight of payoff is minimal and the individual fitness is dominated by the relationship indices gained from the quality of links to neighbors. As a consequence, the changes in the distribution of relationship indices are closely correlated with overall levels of cooperation., Comment: 9 pages, 5 figures

Text-to-SQL generation aims to translate natural language questions into SQL statements. In Text-to-SQL based on large language models, schema linking is a widely adopted strategy to streamline the input for LLMs by selecting only relevant schema elements, therefore reducing noise and computational overhead. However, schema linking faces risks that require caution, including the potential omission of necessary elements and disruption of database structural integrity. To address these challenges, we propose a novel framework called RSL-SQL that combines bidirectional schema linking, contextual information augmentation, binary selection strategy, and multi-turn self-correction. We improve the recall of pattern linking using forward and backward pruning methods, achieving a strict recall of 94% while reducing the number of input columns by 83%. Furthermore, it hedges the risk by voting between a full mode and a simplified mode enhanced with contextual information. Experiments on the BIRD and Spider benchmarks demonstrate that our approach achieves SOTA execution accuracy among open-source solutions, with 67.2% on BIRD and 87.9% on Spider using GPT-4o. Furthermore, our approach outperforms a series of GPT-4 based Text-to-SQL systems when adopting DeepSeek (much cheaper) with same intact prompts. Extensive analysis and ablation studies confirm the effectiveness of each component in our framework. The codes are available at https://github.com/Laqcce-cao/RSL-SQL.

We introduce anytime constraints to the multi-agent setting with the corresponding solution concept being anytime-constrained equilibrium (ACE). Then, we present a comprehensive theory of anytime-constrained Markov games, which includes (1) a computational characterization of feasible policies, (2) a fixed-parameter tractable algorithm for computing ACE, and (3) a polynomial-time algorithm for approximately computing feasible ACE. Since computing a feasible policy is NP-hard even for two-player zero-sum games, our approximation guarantees are the best possible under worst-case analysis. We also develop the first theory of efficient computation for action-constrained Markov games, which may be of independent interest.

We report the discovery of a peculiar X-ray transient, EP240408a, by Einstein Probe (EP) and follow-up studies made with EP, Swift, NICER, GROND, ATCA and other ground-based multi-wavelength telescopes. The new transient was first detected with Wide-field X-ray Telescope (WXT) on board EP on April 8th, 2024, manifested in an intense yet brief X-ray flare lasting for 12 seconds. The flare reached a peak flux of 3.9x10^(-9) erg/cm2/s in 0.5-4 keV, about 300 times brighter than the underlying X-ray emission detected throughout the observation. Rapid and more precise follow-up observations by EP/FXT, Swift and NICER confirmed the finding of this new transient. Its X-ray spectrum is non-thermal in 0.5-10 keV, with a power-law photon index varying within 1.8-2.5. The X-ray light curve shows a plateau lasting for about 4 days, followed by a steep decay till becoming undetectable about 10 days after the initial detection. Based on its temporal property and constraints from previous EP observations, an unusual timescale in the range of 7-23 days is found for EP240408a, which is intermediate between the commonly found fast and long-term transients. No counterparts have been found in optical and near-infrared, with the earliest observation at 17 hours after the initial X-ray detection, suggestive of intrinsically weak emission in these bands. We demonstrate that the remarkable properties of EP240408a are inconsistent with any of the transient types known so far, by comparison with, in particular, jetted tidal disruption events, gamma-ray bursts, X-ray binaries and fast blue optical transients. The nature of EP240408a thus remains an enigma. We suggest that EP240408a may represent a new type of transients with intermediate timescales of the order of about 10 days. The detection and follow-ups of more of such objects are essential for revealing their origin., Comment: 25 pages, 11 figures

The Lobster Eye Imager for Astronomy (LEIA) detected a new X-ray transient on 2022 November 7, identified as a superflare event occurring on a nearby K-type giant star HD 251108. The flux increase was also detected in follow-up observations at X-ray, UV and optical wavelengths. The flare lasted for about 40 days in soft X-ray observations, reaching a peak luminosity of ~1.1 * 10^34 erg/s in 0.5-4.0 keV, which is roughly 60 times the quiescent luminosity. Optical brightening was observed for only one night. The X-ray light curve is well described by a double fast rise and exponential decay model, attributed to the cooling process of a loop arcade structure formed subsequent to the initial large loop with a half-length of ~1.9 * 10^12 cm. Time-resolved X-ray spectra were fitted by a four-temperature apec model (with three components being the quiescent background), showing significant evolution of plasma temperature and emission measure over time. The estimated energy released in the LEIA band is ~3 * 10^39 erg, suggesting that this is likely the most energetic X-ray stellar flare with the longest duration detected to date., Comment: published in ApJ, 15 pages, 7 figures, 8 tables

In recent advancements, multimodal large language models (MLLMs) have been fine-tuned on specific medical image datasets to address medical visual question answering (Med-VQA) tasks. However, this common approach of task-specific fine-tuning is costly and necessitates separate models for each downstream task, limiting the exploration of zero-shot capabilities. In this paper, we introduce MC-CoT, a modular cross-modal collaboration Chain-of-Thought (CoT) framework designed to enhance the zero-shot performance of MLLMs in Med-VQA by leveraging large language models (LLMs). MC-CoT improves reasoning and information extraction by integrating medical knowledge and task-specific guidance, where LLM provides various complex medical reasoning chains and MLLM provides various observations of medical images based on instructions of the LLM. Our experiments on datasets such as SLAKE, VQA-RAD, and PATH-VQA show that MC-CoT surpasses standalone MLLMs and various multimodality CoT frameworks in recall rate and accuracy. These findings highlight the importance of incorporating background information and detailed guidance in addressing complex zero-shot Med-VQA tasks., Comment: 21 pages, 14 figures, 6 tables

With the advent of Large Language Models (LLMs), generating rule-based data for real-world applications has become more accessible. Due to the inherent ambiguity of natural language and the complexity of rule sets, especially in long contexts, LLMs often struggle to follow all specified rules, frequently omitting at least one. To enhance the reasoning and understanding of LLMs on long and complex contexts, we propose a novel prompting strategy Multi-Lingual Prompt, namely MLPrompt, which automatically translates the error-prone rule that an LLM struggles to follow into another language, thus drawing greater attention to it. Experimental results on public datasets across various tasks have shown MLPrompt can outperform state-of-the-art prompting methods such as Chain of Thought, Tree of Thought, and Self-Consistency. Additionally, we introduce a framework integrating MLPrompt with an auto-checking mechanism for structured data generation, with a specific case study in text-to-MIP instances. Further, we extend the proposed framework for text-to-SQL to demonstrate its generation ability towards structured data synthesis.

We introduce CHARTOM, a visual theory-of-mind benchmark for multimodal large language models. CHARTOM consists of specially designed data visualizing charts. Given a chart, a language model needs to not only correctly comprehend the chart (the FACT question) but also judge if the chart will be misleading to a human reader (the MIND question). Both questions have significant societal benefits. We detail the construction of the CHARTOM benchmark including its calibration on human performance.

This paper demonstrates the impact of state-of-the-art instrumental calibration techniques on the precision of arrival times obtained from 9.6 years of observations of millisecond pulsars using the Murriyang 64-m CSIRO Parkes Radio Telescope. Our study focuses on 21-cm observations of 25 high-priority pulsars that are regularly observed as part of the Parkes Pulsar Timing Array (PPTA) project, including those predicted to be the most susceptible to calibration errors. We employ Measurement Equation Template Matching (METM) for instrumental calibration and Matrix Template Matching (MTM) for arrival time estimation, resulting in significantly improved timing residuals with up to a sixfold reduction in white noise compared to arrival times estimated using Scalar Template Matching and conventional calibration based on the Ideal Feed Assumption. The median relative reduction in white noise is 33 percent, and the maximum absolute reduction is 4.5 microseconds. For PSR J0437-4715, METM and MTM reduce the best-fit power-law amplitude (2.7 sigma) and spectral index (1.7 sigma) of the red noise in the arrival time residuals, which can can be tentatively interpreted as mitigation of 1/f noise due to otherwise unmodeled steps in polarimetric response. These findings demonstrate the potential to directly enhance the sensitivity of pulsar timing array experiments through more accurate methods of instrumental calibration and arrival time estimation., Comment: 30 pages, 9 figures, accepted for publication in The Astrophysical Journal

Federated learning has emerged as a promising paradigm for collaborative model training while preserving data privacy. However, recent studies have shown that it is vulnerable to various privacy attacks, such as data reconstruction attacks. In this paper, we provide a theoretical analysis of privacy leakage in federated learning from two perspectives: linear algebra and optimization theory. From the linear algebra perspective, we prove that when the Jacobian matrix of the batch data is not full rank, there exist different batches of data that produce the same model update, thereby ensuring a level of privacy. We derive a sufficient condition on the batch size to prevent data reconstruction attacks. From the optimization theory perspective, we establish an upper bound on the privacy leakage in terms of the batch size, the distortion extent, and several other factors. Our analysis provides insights into the relationship between privacy leakage and various aspects of federated learning, offering a theoretical foundation for designing privacy-preserving federated learning algorithms.

We study a collaborative learning problem where $m$ agents estimate a vector $\mu\in\mathbb{R}^d$ by collecting samples from normal distributions, with each agent $i$ incurring a cost $c_{i,k} \in (0, \infty]$ to sample from the $k^{\text{th}}$ distribution $\mathcal{N}(\mu_k, \sigma^2)$. Instead of working on their own, agents can collect data that is cheap to them, and share it with others in exchange for data that is expensive or even inaccessible to them, thereby simultaneously reducing data collection costs and estimation error. However, when agents have different collection costs, we need to first decide how to fairly divide the work of data collection so as to benefit all agents. Moreover, in naive sharing protocols, strategic agents may under-collect and/or fabricate data, leading to socially undesirable outcomes. Our mechanism addresses these challenges by combining ideas from cooperative and non-cooperative game theory. We use ideas from axiomatic bargaining to divide the cost of data collection. Given such a solution, we develop a Nash incentive-compatible (NIC) mechanism to enforce truthful reporting. We achieve a $\mathcal{O}(\sqrt{m})$ approximation to the minimum social penalty (sum of agent estimation errors and data collection costs) in the worst case, and a $\mathcal{O}(1)$ approximation under favorable conditions. We complement this with a hardness result, showing that $\Omega(\sqrt{m})$ is unavoidable in any NIC mechanism.

The observation of neutron stars enables the otherwise impossible study of fundamental physical processes. The timing of binary radio pulsars is particularly powerful, as it enables precise characterization of their (three-dimensional) positions and orbits. PSR~J0437$-$4715 is an important millisecond pulsar for timing array experiments and is also a primary target for the Neutron Star Interior Composition Explorer (NICER). The main aim of the NICER mission is to constrain the neutron star equation of state by inferring the compactness ($M_p/R$) of the star. Direct measurements of the mass $M_p$ from pulsar timing therefore substantially improve constraints on the radius $R$ and the equation of state. Here we use observations spanning 26 years from Murriyang, the 64-m Parkes radio telescope, to improve the timing model for this pulsar. Among the new precise measurements are the pulsar mass $M_p=1.418\pm 0.044$ $M_{\odot}$, distance $D=156.96 \pm 0.11$ pc, and orbital inclination angle $i=137.506 \pm 0.016^\circ$, which can be used to inform the X-ray pulse profile models inferred from NICER observations. We demonstrate that these results are consistent between multiple data sets from the Parkes Pulsar Timing Array (PPTA), each modeled with different noise assumptions. Using the longest available PPTA data set, we measure an apparent second derivative of the pulsar spin frequency and discuss how this can be explained either by kinematic effects due to the proper motion and radial velocity of the pulsar or excess low-frequency noise such as a gravitational-wave background., Comment: 13 pages, 3 figures, accepted for publication in Astrophysical Journal Letters

In this paper, we first give an introduction to the theoretical basis of the privacy-utility equilibrium in federated learning based on Bayesian privacy definitions and total variation distance privacy definitions. We then present the \textit{Learn-to-Distort-Data} framework, which provides a principled approach to navigate the privacy-utility equilibrium by explicitly modeling the distortion introduced by the privacy-preserving mechanism as a learnable variable and optimizing it jointly with the model parameters. We demonstrate the applicability of our framework to a variety of privacy-preserving mechanisms on the basis of data distortion and highlight its connections to related areas such as adversarial training, input robustness, and unlearnable examples. These connections enable leveraging techniques from these areas to design effective algorithms for privacy-utility equilibrium in federated learning under the \textit{Learn-to-Distort-Data} framework.

We study security threats to Markov games due to information asymmetry and misinformation. We consider an attacker player who can spread misinformation about its reward function to influence the robust victim player's behavior. Given a fixed fake reward function, we derive the victim's policy under worst-case rationality and present polynomial-time algorithms to compute the attacker's optimal worst-case policy based on linear programming and backward induction. Then, we provide an efficient inception ("planting an idea in someone's mind") attack algorithm to find the optimal fake reward function within a restricted set of reward functions with dominant strategies. Importantly, our methods exploit the universal assumption of rationality to compute attacks efficiently. Thus, our work exposes a security vulnerability arising from standard game assumptions under misinformation., Comment: Accepted to Reinforcement Learning Conference (RLC) 2024

In social dilemmas, most interactions are transient and susceptible to restructuring, leading to continuous changes in social networks over time. Typically, agents assess the rewards of their current interactions and adjust their connections to optimize outcomes. In this paper, we introduce an adaptive network model in the snowdrift game to examine dynamic levels of cooperation and network topology, involving the potential for both the termination of existing connections and the establishment of new ones. In particular, we define the agent's asymmetric disassociation tendency toward their neighbors, which fundamentally determines the probability of edge dismantlement. The mechanism allows agents to selectively sever and rewire their connections to alternative individuals to refine partnerships. Our findings reveal that adaptive networks are particularly effective in promoting a robust evolution toward states of either pure cooperation or complete defection, especially under conditions of extreme cost-benefit ratios, as compared to static network models. Moreover, the dynamic restructuring of connections and the distribution of network degrees among agents are closely linked to the levels of cooperation in stationary states. Specifically, cooperators tend to seek broader neighborhoods when confronted with the invasion of multiple defectors., Comment: 12 pages, 6 figures

Large Language Models (LLMs) have training corpora containing large amounts of program code, greatly improving the model's code comprehension and generation capabilities. However, sound comprehensive research on detecting program vulnerabilities, a more specific task related to code, and evaluating the performance of LLMs in this more specialized scenario is still lacking. To address common challenges in vulnerability analysis, our study introduces a new benchmark, VulDetectBench, specifically designed to assess the vulnerability detection capabilities of LLMs. The benchmark comprehensively evaluates LLM's ability to identify, classify, and locate vulnerabilities through five tasks of increasing difficulty. We evaluate the performance of 17 models (both open- and closed-source) and find that while existing models can achieve over 80% accuracy on tasks related to vulnerability identification and classification, they still fall short on specific, more detailed vulnerability analysis tasks, with less than 30% accuracy, making it difficult to provide valuable auxiliary information for professional vulnerability mining. Our benchmark effectively evaluates the capabilities of various LLMs at different levels in the specific task of vulnerability detection, providing a foundation for future research and improvements in this critical area of code security. VulDetectBench is publicly available at https://github.com/Sweetaroo/VulDetectBench.

Individuals and businesses have been significantly benefited by Large Language Models (LLMs) including PaLM, Gemini and ChatGPT in various ways. For example, LLMs enhance productivity, reduce costs, and enable us to focus on more valuable tasks. Furthermore, LLMs possess the capacity to sift through extensive datasets, uncover underlying patterns, and furnish critical insights that propel the frontiers of technology and science. However, LLMs also pose privacy concerns. Users' interactions with LLMs may expose their sensitive personal or company information. A lack of robust privacy safeguards and legal frameworks could permit the unwarranted intrusion or improper handling of individual data, thereby risking infringements of privacy and the theft of personal identities. To ensure privacy, it is essential to minimize the dependency between shared prompts and private information. Various randomization approaches have been proposed to protect prompts' privacy, but they may incur utility loss compared to unprotected LLMs prompting. Therefore, it is essential to evaluate the balance between the risk of privacy leakage and loss of utility when conducting effective protection mechanisms. The current study develops a framework for inferring privacy-protected Large Language Models (LLMs) and lays down a solid theoretical basis for examining the interplay between privacy preservation and utility. The core insight is encapsulated within a theorem that is called as the NFL (abbreviation of the word No-Free-Lunch) Theorem.

This work presents a tuning-free semantic segmentation framework based on classifying SAM masks by CLIP, which is universally applicable to various types of supervision. Initially, we utilize CLIP's zero-shot classification ability to generate pseudo-labels or perform open-vocabulary segmentation. However, the misalignment between mask and CLIP text embeddings leads to suboptimal results. To address this issue, we propose discrimination-bias aligned CLIP to closely align mask and text embedding, offering an overhead-free performance gain. We then construct a global-local consistent classifier to classify SAM masks, which reveals the intrinsic structure of high-quality embeddings produced by DBA-CLIP and demonstrates robustness against noisy pseudo-labels. Extensive experiments validate the efficiency and effectiveness of our method, and we achieve state-of-the-art (SOTA) or competitive performance across various datasets and supervision types.

Weakly supervised multimodal violence detection aims to learn a violence detection model by leveraging multiple modalities such as RGB, optical flow, and audio, while only video-level annotations are available. In the pursuit of effective multimodal violence detection (MVD), information redundancy, modality imbalance, and modality asynchrony are identified as three key challenges. In this work, we propose a new weakly supervised MVD method that explicitly addresses these challenges. Specifically, we introduce a multi-scale bottleneck transformer (MSBT) based fusion module that employs a reduced number of bottleneck tokens to gradually condense information and fuse each pair of modalities and utilizes a bottleneck token-based weighting scheme to highlight more important fused features. Furthermore, we propose a temporal consistency contrast loss to semantically align pairwise fused features. Experiments on the largest-scale XD-Violence dataset demonstrate that the proposed method achieves state-of-the-art performance. Code is available at https://github.com/shengyangsun/MSBT., Comment: Accepted by ICME 2024

We present DeepSeek-V2, a strong Mixture-of-Experts (MoE) language model characterized by economical training and efficient inference. It comprises 236B total parameters, of which 21B are activated for each token, and supports a context length of 128K tokens. DeepSeek-V2 adopts innovative architectures including Multi-head Latent Attention (MLA) and DeepSeekMoE. MLA guarantees efficient inference through significantly compressing the Key-Value (KV) cache into a latent vector, while DeepSeekMoE enables training strong models at an economical cost through sparse computation. Compared with DeepSeek 67B, DeepSeek-V2 achieves significantly stronger performance, and meanwhile saves 42.5% of training costs, reduces the KV cache by 93.3%, and boosts the maximum generation throughput to 5.76 times. We pretrain DeepSeek-V2 on a high-quality and multi-source corpus consisting of 8.1T tokens, and further perform Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) to fully unlock its potential. Evaluation results show that, even with only 21B activated parameters, DeepSeek-V2 and its chat versions still achieve top-tier performance among open-source models.

Proteins are essential to life's processes, underpinning evolution and diversity. Advances in sequencing technology have revealed millions of proteins, underscoring the need for sophisticated pre-trained protein models for biological analysis and AI development. Facebook's ESM2, the most advanced protein language model to date, leverages a masked prediction task for unsupervised learning, crafting amino acid representations with notable biochemical accuracy. Yet, it lacks in delivering functional protein insights, signaling an opportunity for enhancing representation quality.Our study addresses this gap by incorporating protein family classification into ESM2's training.This approach, augmented with Community Propagation-Based Clustering Algorithm, improves global protein representations, while a contextual prediction task fine-tunes local amino acid accuracy. Significantly, our model achieved state-of-the-art results in several downstream experiments, demonstrating the power of combining global and local methodologies to substantially boost protein representation quality.

We give a direct proof of the following known result: the Grothendieck group of a triangulated category with a silting subcategory is isomorphic to the split Grothendieck group of the silting subcategory. Moreover, we obtain its cluster-tilting analogue., Comment: added a new section to discuss the cluster-tilting analogue