Gradient_Descent
-
Energy-Based Transformers are Scalable Learners and Thinkers
2025-07-02 19:17:29Alexi Gladstone, Ganesh Nanduru, Md Mofijul Islam, Peixuan Han, Hyeonjeong Ha, Aman Chadha, Yilun Du, Heng Ji, Jundong Li, Tariq IqbalAbstract
Inference-time computation techniques, analogous to human System 2 Thinking, have recently become popular for improving model performances. However, most existing approaches suffer from several limitations: they are modality-specific (e.g., working only in text), problem-specific (e.g., verifiable domains like math and coding), or require additional supervision/training on top of unsupervised pretraining (e.g., verifiers or verifiable rewards). In this paper, we ask the question "Is it possible to generalize these System 2 Thinking approaches, and develop models that learn to think solely from unsupervised learning?" Interestingly, we find the answer is yes, by learning to explicitly verify the compatibility between inputs and candidate-predictions, and then re-framing prediction problems as optimization with respect to this verifier. Specifically, we train Energy-Based Transformers (EBTs) -- a new class of Energy-Based Models (EBMs) -- to assign an energy value to every input and candidate-prediction pair, enabling predictions through gradient descent-based energy minimization until convergence. Across both discrete (text) and continuous (visual) modalities, we find EBTs scale faster than the dominant Transformer++ approach during training, achieving an up to 35% higher scaling rate with respect to data, batch size, parameters, FLOPs, and depth. During inference, EBTs improve performance with System 2 Thinking by 29% more than the Transformer++ on language tasks, and EBTs outperform Diffusion Transformers on image denoising while using fewer forward passes. Further, we find that EBTs achieve better results than existing models on most downstream tasks given the same or worse pretraining performance, suggesting that EBTs generalize better than existing approaches. Consequently, EBTs are a promising new paradigm for scaling both the learning and thinking capabilities of models.
Abstract (translated)
类似于人类的系统2思维，推理时的计算技术最近因其在提升模型性能方面的潜力而变得流行。然而，大多数现有的方法存在若干局限性：它们只适用于特定模式（例如仅限于文本），针对特定问题领域（如数学和编程验证域），或者需要在无监督预训练基础上额外的监督或训练（例如校验器或可验证奖励）。在这篇论文中，我们提出了一个问题：“是否有可能将这些系统2思维方法进行泛化，并开发出仅仅通过无监督学习就能学会思考的模型？”令人惊讶的是，我们发现答案是肯定的。具体而言，我们的研究发现通过学习明确地验证输入与候选预测之间的兼容性，并重新框架为基于校验器的优化问题，可以实现这一目标。也就是说，我们训练了一种新型的能量基础模型（EBMs），即能量基础变换器（EBTs）来给每个输入和候选预测对分配一个能量值，使通过基于梯度下降的最小化直至收敛来进行预测成为可能。 在离散（文本）和连续（视觉）模式上，我们发现与目前主导的方法Transformer++相比，在训练过程中，EBTs的学习速率更快。具体来说，在数据量、批次大小、参数数量、FLOPs以及模型深度等方面，EBTs的扩展速度比Transformer++快35%以上。在推理阶段，相对于Transformer++方法，EBTs在语言任务上的系统2思维性能提高了29%，并且在图像去噪上优于扩散变换器，同时使用了更少的前向传递次数。此外，我们还发现，在给定相同或较差预训练表现的情况下，EBTs在大多数下游任务中的表现比现有的模型更好，这表明EBTs具有更好的泛化能力。 因此，能量基础变换器（EBTs）为增强模型的学习和推理能力提供了一种有前景的新范式。
URL
https://arxiv.org/abs/2507.02092
PDF
-
Gradient-Adaptive Policy Optimization: Towards Multi-Objective Alignment of Large Language Models
2025-07-02 17:25:26Chengao Li, Hanyu Zhang, Yunkun Xu, Hongyan Xue, Xiang Ao, Qing HeAbstract
Reinforcement Learning from Human Feedback (RLHF) has emerged as a powerful technique for aligning large language models (LLMs) with human preferences. However, effectively aligning LLMs with diverse human preferences remains a significant challenge, particularly when they are conflict. To address this issue, we frame human value alignment as a multi-objective optimization problem, aiming to maximize a set of potentially conflicting objectives. We introduce Gradient-Adaptive Policy Optimization (GAPO), a novel fine-tuning paradigm that employs multiple-gradient descent to align LLMs with diverse preference distributions. GAPO adaptively rescales the gradients for each objective to determine an update direction that optimally balances the trade-offs between objectives. Additionally, we introduce P-GAPO, which incorporates user preferences across different objectives and achieves Pareto solutions that better align with the user's specific needs. Our theoretical analysis demonstrates that GAPO converges towards a Pareto optimal solution for multiple objectives. Empirical results on Mistral-7B show that GAPO outperforms current state-of-the-art methods, achieving superior performance in both helpfulness and harmlessness.
Abstract (translated)
从人类反馈中的强化学习（RLHF）已成为将大型语言模型（LLM）与人类偏好对齐的强大技术。然而，有效地使LLM与多样化的甚至是冲突的人类偏好相一致仍是一个重大挑战。为了解决这个问题，我们将人类价值观的对齐问题构架成一个多目标优化问题，旨在最大化一组潜在冲突的目标。我们引入了一种新型微调范式——梯度自适应策略优化（GAPO），它利用多梯度下降来使LLM与多种偏好分布相一致。GAPO能够根据每个目标自适应地调整梯度的尺度，以确定一个最优平衡各目标间权衡的更新方向。此外，我们还引入了P-GAPO，该方法结合不同目标上的用户偏好，并实现帕累托解，更好地满足用户的特定需求。理论分析表明，GAPO可以朝着多目标问题的帕累托最优解收敛。在Mistral-7B模型上的实证结果显示，GAPO超越了当前最先进的方法，在有用性和无害性方面表现出色。
URL
https://arxiv.org/abs/2507.01915
PDF
-
Description of the Training Process of Neural Networks via Ergodic Theorem : Ghost nodes
Abstract
Recent studies have proposed interpreting the training process from an ergodic perspective. Building on this foundation we present a unified framework for understanding and accelerating the training of deep neural networks via stochastic gradient descent. By analyzing the geometric landscape of the objective function we introduce a practical diagnostic, the running estimate of the largest Lyapunov exponent, which provably distinguishes genuine convergence toward stable minimizers from mere statistical stabilization near saddle points. We then propose a ghost category extension for standard classifiers that adds auxiliary ghost output nodes so the model gains extra descent directions that open a lateral corridor around narrow loss barriers and enable the optimizer to bypass poor basins during the early training phase. We show that this extension strictly reduces approximation error and that after sufficient convergence the ghost dimensions collapse and the extended model's invariant law coincides with that of the original and there exists a path in the enlarged parameter space along which the total loss does not increase while the original loss decreases by an arbitrary margin. Taken together these results provide a principled architecture level intervention that accelerates early stage trainability while preserving asymptotic behavior.
Abstract (translated)
近期的研究提出从遍历性视角来解释训练过程。在此基础上，我们提出了一个统一的框架，通过随机梯度下降法理解并加速深度神经网络的训练。通过对目标函数几何景观的分析，我们引入了一个实用的诊断工具——运行中的最大Lyapunov指数估计值，该方法可以证明区分真正的向稳定极小值点收敛和仅仅在鞍点附近的统计稳定。 接着，我们提出了一种标准分类器的幽灵类别扩展方法，即添加辅助的幽灵输出节点。这样模型可以获得额外的下降方向，在狭窄损失障碍物周围打开横向通道，并使优化器能够在训练初期绕过不良盆地。 我们展示了这种扩展能严格减少近似误差，并且在充分收敛后，幽灵维度会坍缩，扩展后的模型不变性定律与原模型一致，同时在扩大的参数空间中存在一条路径，在这条路径上总损失不会增加而原始损失可以任意减小。 综上所述，这些结果提供了一种基于架构的干预方法，可以在保持渐近行为的同时加速早期训练阶段。
URL
https://arxiv.org/abs/2507.01003
PDF
-
Data Uniformity Improves Training Efficiency and More, with a Convergence Framework Beyond the NTK Regime
2025-06-30 17:58:30Yuqing Wang, Shangding GuAbstract
Data selection plays a crucial role in data-driven decision-making, including in large language models (LLMs), and is typically task-dependent. Properties such as data quality and diversity have been extensively studied and are known to enhance model performance. However, it remains unclear whether there exist other quantitative and general principles of data selection that can consistently improve performance, especially for complex tasks with limited prior knowledge. In this paper, we demonstrate that selecting more uniformly distributed data can improve training efficiency while enhancing performance. Specifically, we establish that more uniform (less biased) distribution leads to a larger minimum pairwise distance between data points, denoted by $h_{\min}$, and prove that a smaller $h_{\min}$ can slow down the training dynamics of gradient descent (GD). Moreover, we theoretically show that the approximation error of neural networks decreases as $h_{\min}$ increases. Our analysis introduces a convergence framework for GD beyond the Neural Tangent Kernel (NTK) regime, applicable to a broad class of architectures, including transformers, without requiring Lipschitz smoothness. This framework further provides theoretical justification for the use of residual connections and function compositions in deep neural architectures. In the end, we conduct comprehensive experiments for supervised fine-tuning across various settings, including different optimization strategies, model sizes, and training datasets. The results consistently demonstrate that selecting data by maximizing pairwise distance significantly accelerates training and achieves comparable or better performance in LLMs across diverse datasets. Code and Datasets are available at the link: this https URL.
Abstract (translated)
数据选择在基于数据的决策制定中扮演着关键角色，特别是在大型语言模型（LLMs）中，并且通常是任务相关的。诸如数据质量和多样性的属性已经被广泛研究并且被证明可以提高模型性能。然而，是否存在其他能够一致提升性能的数据选择定量和通用原则，尤其是对于那些先验知识有限的复杂任务来说，这一点尚不清楚。 在本文中，我们展示了选择更均匀分布的数据可以在提高训练效率的同时增强性能。具体而言，我们确立了数据点之间最小成对距离（记为 $h_{\min}$）随分布更加均匀而增加，并证明较小的 $h_{\min}$ 可以减缓梯度下降 (GD) 的训练动态。此外，理论上显示神经网络的近似误差随着 $h_{\min}$ 的增大而减少。 我们的分析引入了一个超越神经切片核（NTK）框架的收敛性理论框架，适用于包括变压器在内的广泛架构类别，并且不需要利普希茨平滑度。此框架进一步提供了残差连接和函数组合在深度神经网络结构中的使用依据。 最终，我们在各种设置下进行了全面的监督微调实验，包括不同的优化策略、模型大小和训练数据集。结果一致表明，在大型语言模型中，通过最大化成对距离来选择数据可以显著加速训练，并且在不同数据集上可获得相当或更好的性能。 代码和数据集可以在以下链接获取：[此URL]（请将 [此URL] 替换为实际的链接地址）。
URL
https://arxiv.org/abs/2506.24120
PDF
-
SPI-BoTER: Error Compensation for Industrial Robots via Sparse Attention Masking and Hybrid Loss with Spatial-Physical Information
2025-06-28 07:00:59Xuao Hou, Yongquan Jia, Shijin Zhang, Yuqiang WuAbstract
The widespread application of industrial robots in fields such as cutting and welding has imposed increasingly stringent requirements on the trajectory accuracy of end-effectors. However, current error compensation methods face several critical challenges, including overly simplified mechanism modeling, a lack of physical consistency in data-driven approaches, and substantial data requirements. These issues make it difficult to achieve both high accuracy and strong generalization simultaneously. To address these challenges, this paper proposes a Spatial-Physical Informed Attention Residual Network (SPI-BoTER). This method integrates the kinematic equations of the robotic manipulator with a Transformer architecture enhanced by sparse self-attention masks. A parameter-adaptive hybrid loss function incorporating spatial and physical information is employed to iteratively optimize the network during training, enabling high-precision error compensation under small-sample conditions. Additionally, inverse joint angle compensation is performed using a gradient descent-based optimization method. Experimental results on a small-sample dataset from a UR5 robotic arm (724 samples, with a train:test:validation split of 8:1:1) demonstrate the superior performance of the proposed method. It achieves a 3D absolute positioning error of 0.2515 mm with a standard deviation of 0.15 mm, representing a 35.16\% reduction in error compared to conventional deep neural network (DNN) methods. Furthermore, the inverse angle compensation algorithm converges to an accuracy of 0.01 mm within an average of 147 iterations. This study presents a solution that combines physical interpretability with data adaptability for high-precision control of industrial robots, offering promising potential for the reliable execution of precision tasks in intelligent manufacturing.
Abstract (translated)
工业机器人在切割和焊接等领域广泛应用，对末端执行器的轨迹精度提出了越来越严格的要求。然而，目前的误差补偿方法面临着几个关键挑战，包括过于简化的机制建模、数据驱动方法中物理一致性的缺乏以及大量数据的需求。这些问题使得同时实现高精度和强泛化能力变得困难。为了解决这些挑战，本文提出了一种基于空间-物理信息注意残差网络（SPI-BoTER）的方法。该方法将机器人操作臂的运动学方程与增强型稀疏自注意力掩码的Transformer架构相结合，并在训练过程中采用结合了空间和物理信息的参数自适应混合损失函数进行迭代优化，从而能够在小样本条件下实现高精度误差补偿。此外，还使用基于梯度下降的优化方法来进行反向关节角度补偿。 实验结果表明，在一个UR5机器人手臂的小样本数据集（724个样本，训练、测试和验证的比例为8:1:1）上，所提出的方法表现出优越性能。它实现了0.2515毫米的三维绝对定位误差，标准差为0.15毫米，相较于传统的深度神经网络方法，误差降低了35.16%。此外，反向角度补偿算法在平均147次迭代后收敛到0.01毫米精度。 这项研究提供了一个结合物理可解释性和数据适应性的解决方案，用于实现工业机器人的高精度控制，在智能制造中可靠地执行精密任务方面展现了巨大的潜力。
URL
https://arxiv.org/abs/2506.22788
PDF
-
Curriculum-Guided Antifragile Reinforcement Learning for Secure UAV Deconfliction under Observation-Space Attacks
2025-06-26 10:10:41Deepak Kumar Panda, Adolfo Perrusquia, Weisi GuoAbstract
Reinforcement learning (RL) policies deployed in safety-critical systems, such as unmanned aerial vehicle (UAV) navigation in dynamic airspace, are vulnerable to out-ofdistribution (OOD) adversarial attacks in the observation space. These attacks induce distributional shifts that significantly degrade value estimation, leading to unsafe or suboptimal decision making rendering the existing policy fragile. To address this vulnerability, we propose an antifragile RL framework designed to adapt against curriculum of incremental adversarial perturbations. The framework introduces a simulated attacker which incrementally increases the strength of observation-space perturbations which enables the RL agent to adapt and generalize across a wider range of OOD observations and anticipate previously unseen attacks. We begin with a theoretical characterization of fragility, formally defining catastrophic forgetting as a monotonic divergence in value function distributions with increasing perturbation strength. Building on this, we define antifragility as the boundedness of such value shifts and derive adaptation conditions under which forgetting is stabilized. Our method enforces these bounds through iterative expert-guided critic alignment using Wasserstein distance minimization across incrementally perturbed observations. We empirically evaluate the approach in a UAV deconfliction scenario involving dynamic 3D obstacles. Results show that the antifragile policy consistently outperforms standard and robust RL baselines when subjected to both projected gradient descent (PGD) and GPS spoofing attacks, achieving up to 15% higher cumulative reward and over 30% fewer conflict events. These findings demonstrate the practical and theoretical viability of antifragile reinforcement learning for secure and resilient decision-making in environments with evolving threat scenarios.
Abstract (translated)
强化学习（RL）策略在无人驾驶航空器（UAV）导航等安全关键系统中的应用，面对动态空域中观察空间的分布外（OOD）对抗攻击时表现出脆弱性。这些攻击导致价值估计出现显著偏差，从而引发不安全或次优决策，使现有策略变得不稳定。为了应对这一漏洞，我们提出了一种抗脆性的RL框架，该框架能够适应增量式对抗扰动的课程训练。该框架引入了一个模拟攻击者，逐步增加观察空间中的干扰强度，从而使强化学习代理能够在更广泛的OOD观察中进行自适应和泛化，并且可以预测以前未见过的攻击。 我们从理论上对脆弱性进行了描述，正式定义了灾难性遗忘为随着扰动强度增强价值函数分布之间的单调发散。在此基础上，我们将抗脆性定义为这样的价值转移是有限度的，并推导出在哪些适应条件下可以使这种遗忘稳定下来。我们的方法通过使用Wasserstein距离最小化来进行迭代专家指导的批评对齐来强制执行这些限制，该过程是在逐步受到扰动的观察中进行的。 我们在涉及动态3D障碍物规避的UAV场景中进行了实验评估。结果显示，在面对投影梯度下降（PGD）和GPS欺骗攻击时，抗脆性策略在标准RL基准线的基础上累计奖励提高了高达15%，冲突事件减少了超过30%。这些发现证明了抗脆性强化学习对于具有不断变化威胁环境中的安全和弹性决策制定的实用性和理论可行性。 总之，该研究为强化学习算法的安全性和稳健性提供了一种新的方法，并展示了在面对动态且复杂的OOD攻击时，通过适应训练来提高策略鲁棒性的潜力。
URL
https://arxiv.org/abs/2506.21129
PDF
-
Robust Policy Switching for Antifragile Reinforcement Learning for UAV Deconfliction in Adversarial Environments
2025-06-26 10:06:29Deepak Kumar Panda, Weisi GuoAbstract
The increasing automation of navigation for unmanned aerial vehicles (UAVs) has exposed them to adversarial attacks that exploit vulnerabilities in reinforcement learning (RL) through sensor manipulation. Although existing robust RL methods aim to mitigate such threats, their effectiveness has limited generalization to out-of-distribution shifts from the optimal value distribution, as they are primarily designed to handle fixed perturbation. To address this limitation, this paper introduces an antifragile RL framework that enhances adaptability to broader distributional shifts by incorporating a switching mechanism based on discounted Thompson sampling (DTS). This mechanism dynamically selects among multiple robust policies to minimize adversarially induced state-action-value distribution shifts. The proposed approach first derives a diverse ensemble of action robust policies by accounting for a range of perturbations in the policy space. These policies are then modeled as a multiarmed bandit (MAB) problem, where DTS optimally selects policies in response to nonstationary Bernoulli rewards, effectively adapting to evolving adversarial strategies. Theoretical framework has also been provided where by optimizing the DTS to minimize the overall regrets due to distributional shift, results in effective adaptation against unseen adversarial attacks thus inducing antifragility. Extensive numerical simulations validate the effectiveness of the proposed framework in complex navigation environments with multiple dynamic three-dimensional obstacles and with stronger projected gradient descent (PGD) and spoofing attacks. Compared to conventional robust, non-adaptive RL methods, the antifragile approach achieves superior performance, demonstrating shorter navigation path lengths and a higher rate of conflict-free navigation trajectories compared to existing robust RL techniques
Abstract (translated)
无人驾驶飞机（UAV）导航的自动化程度提高，使其更容易受到通过操纵传感器来利用强化学习（RL）漏洞的对抗性攻击。尽管现有的鲁棒RL方法旨在减轻此类威胁，但由于它们主要设计用于处理固定扰动，因此在面对偏离最优价值分布的变化时，其效果有限且泛化能力不足。为了克服这一限制，本文提出了一种基于折扣Thompson采样（DTS）切换机制的抗脆弱性RL框架，以增强对更广泛分布变化的适应性。此机制通过动态选择多个鲁棒策略来最小化对抗性引起的状态-动作值分布的变化。 所提出的这种方法首先通过对策略空间中一系列扰动进行建模，导出一组多样化的行动鲁棒策略集合。然后将这些策略作为多臂赌博机（MAB）问题进行处理，在此过程中使用DTS根据非平稳的伯努利奖励来优化选择策略，从而有效适应不断变化的对抗性策略。 此外，本文还提供了一个理论框架，通过优化DTS以最小化由于分布变化导致的整体遗憾值，使得该方法能够有效地对抗未见过的对抗性攻击，并因此具有抗脆弱性。大量的数值模拟验证了所提出的框架在复杂导航环境中的有效性，包括存在多个动态三维障碍物的情况，以及更强烈的投影梯度下降（PGD）和欺骗攻击的情况下。 与传统的鲁棒、非适应性的RL方法相比，该抗脆弱性方法展示了更优的性能：其生成的导航路径更短，并且冲突自由的导航轨迹比例更高。这表明，在面对复杂和动态变化的对抗环境中，所提出的框架能够更好地保护UAV的安全性和稳定性。
URL
https://arxiv.org/abs/2506.21127
PDF
-
Can Gradient Descent Simulate Prompting?
2025-06-26 04:06:20Eric Zhang, Leshem Choshen, Jacob AndreasAbstract
There are two primary ways of incorporating new information into a language model (LM): changing its prompt or changing its parameters, e.g. via fine-tuning. Parameter updates incur no long-term storage cost for model changes. However, for many model updates, prompting is significantly more effective: prompted models can generalize robustly from single examples and draw logical inferences that do not occur under standard fine-tuning. Can models be modified so that fine-tuning does emulate prompting? This paper describes a method for meta-training LMs such that gradient updates emulate the effects of conditioning on new information. Our approach uses tools from gradient-based meta-learning but uses an LM's own prompted predictions as targets, eliminating the need for ground-truth labels. Subsequent gradient descent training recovers some (and occasionally all) of prompted model performance -- showing improvement on the ``reversal curse'' tasks, and answering questions about text passages after a single gradient update. These results suggest that, with appropriate initialization, gradient descent can be surprisingly expressive. Our results suggest new avenues for long-context modeling and offer insight into the generalization capabilities of gradient-based learning.
Abstract (translated)
将新信息融入语言模型（LM）主要有两种方式：更改其提示或调整其参数，例如通过微调来实现。参数更新不会产生长期存储成本。然而，在许多模型更新的情况下，使用提示更加有效：经过提示的模型可以从单个示例中稳健地泛化，并得出逻辑推断，而这些在标准微调过程中并不会发生。是否可以修改模型使微调的效果模拟出提示的效果呢？ 本文描述了一种方法，即通过元训练LM的方式使得梯度更新能够模仿对新信息条件化的效果。我们的方法使用基于梯度的元学习工具，但以语言模型自身受到提示后的预测作为目标，从而消除了对真实标签的需求。后续的梯度下降训练可以恢复一些（有时甚至是全部）由提示带来的模型性能——在“逆转诅咒”任务中表现出改善，并且仅通过一次梯度更新即可回答关于文本段落的问题。这些结果表明，在适当的初始化下，梯度下降具有惊人的表达能力。 我们的研究成果为长上下文建模提供了新的途径，并对基于梯度的学习泛化能力提供了见解。
URL
https://arxiv.org/abs/2506.20989
PDF
-
Toward Decision-Oriented Prognostics: An Integrated Estimate-Optimize Framework for Predictive Maintenance
2025-06-24 15:10:15Zhuojun Xie, Adam Abdin, Yiping FangAbstract
Recent research increasingly integrates machine learning (ML) into predictive maintenance (PdM) to reduce operational and maintenance costs in data-rich operational settings. However, uncertainty due to model misspecification continues to limit widespread industrial adoption. This paper proposes a PdM framework in which sensor-driven prognostics inform decision-making under economic trade-offs within a finite decision space. We investigate two key questions: (1) Does higher predictive accuracy necessarily lead to better maintenance decisions? (2) If not, how can the impact of prediction errors on downstream maintenance decisions be mitigated? We first demonstrate that in the traditional estimate-then-optimize (ETO) framework, errors in probabilistic prediction can result in inconsistent and suboptimal maintenance decisions. To address this, we propose an integrated estimate-optimize (IEO) framework that jointly tunes predictive models while directly optimizing for maintenance outcomes. We establish theoretical finite-sample guarantees on decision consistency under standard assumptions. Specifically, we develop a stochastic perturbation gradient descent algorithm suitable for small run-to-failure datasets. Empirical evaluations on a turbofan maintenance case study show that the IEO framework reduces average maintenance regret up to 22% compared to ETO. This study provides a principled approach to managing prediction errors in data-driven PdM. By aligning prognostic model training with maintenance objectives, the IEO framework improves robustness under model misspecification and improves decision quality. The improvement is particularly pronounced when the decision-making policy is misaligned with the decision-maker's target. These findings support more reliable maintenance planning in uncertain operational environments.
Abstract (translated)
最近的研究越来越多地将机器学习（ML）融入预测性维护（PdM），以减少数据密集型操作环境中的运营和维护成本。然而，由于模型规格不匹配导致的不确定性继续限制了其在工业领域的广泛应用。本文提出了一种基于传感器驱动预测的PdM框架，在该框架中，经济权衡下的决策制定在一个有限的决策空间内进行。我们探讨了两个关键问题：（1）更高的预测准确性是否必然会导致更好的维护决策？（2）如果不尽然，如何减轻预测误差对后续维护决策的影响？ 首先，我们展示了在传统的估计然后优化（ETO）框架中，概率预测中的错误可能导致不一致和次优的维护决策。为了解决这个问题，我们提出了一个集成估计-优化（IEO）框架，该框架联合调优预测模型的同时直接针对维护结果进行最优化。我们在标准假设下建立了理论上的有限样本保证，以确保决策的一致性。具体来说，我们开发了一种适合于小规模运行至故障数据集的随机扰动梯度下降算法。 在对涡轮风扇维护案例的研究中，实证评估表明IEO框架相比于ETO减少了高达22%的平均维护遗憾（即未达到最佳维护效果的程度）。这项研究提供了一个基于数据驱动PdM管理预测误差的原则性方法。通过将诊断模型训练与维护目标相结合，IEO框架在模型规格不匹配的情况下提高了鲁棒性和决策质量。这种改进尤其明显于当决策策略与决策者的目标不一致时。 这些发现支持了在不确定的操作环境中进行更可靠的维护规划。
URL
https://arxiv.org/abs/2506.19698
PDF
-
Optimization-Induced Dynamics of Lipschitz Continuity in Neural Networks
2025-06-23 12:49:13R\'ois\'in Luo, James McDermott, Christian Gagn\'e, Qiang Sun, Colm O'RiordanAbstract
Lipschitz continuity characterizes the worst-case sensitivity of neural networks to small input perturbations; yet its dynamics (i.e. temporal evolution) during training remains under-explored. We present a rigorous mathematical framework to model the temporal evolution of Lipschitz continuity during training with stochastic gradient descent (SGD). This framework leverages a system of stochastic differential equations (SDEs) to capture both deterministic and stochastic forces. Our theoretical analysis identifies three principal factors driving the evolution: (i) the projection of gradient flows, induced by the optimization dynamics, onto the operator-norm Jacobian of parameter matrices; (ii) the projection of gradient noise, arising from the randomness in mini-batch sampling, onto the operator-norm Jacobian; and (iii) the projection of the gradient noise onto the operator-norm Hessian of parameter matrices. Furthermore, our theoretical framework sheds light on such as how noisy supervision, parameter initialization, batch size, and mini-batch sampling trajectories, among other factors, shape the evolution of the Lipschitz continuity of neural networks. Our experimental results demonstrate strong agreement between the theoretical implications and the observed behaviors.
Abstract (translated)
李普希茨连续性表征了神经网络对小输入扰动的最坏情况敏感度；然而，在使用随机梯度下降（SGD）进行训练时，其动力学特性（即随时间的变化过程）仍然研究不足。我们提出了一种严格的数学框架来模拟在使用SGD训练过程中李普希茨连续性的随时间变化过程。该框架利用随机微分方程（SDEs）系统捕捉确定性和随机力的影响。 我们的理论分析识别了驱动演变的三个主要因素： (i) 由优化动力学引起的梯度流向参数矩阵操作员范数雅可比矩阵的投影； (ii) 由于小批量采样中的随机性产生的梯度噪声对操作员范数雅可比矩阵的投影； (iii) 梯度噪声向参数矩阵操作员范数海森矩阵的投影。 此外，我们的理论框架还揭示了诸如嘈杂监督、参数初始化、批次大小和小批量采样轨迹等因素如何塑造神经网络李普希茨连续性随时间的变化。实验结果表明，理论推断与观察到的行为之间有很强的一致性。
URL
https://arxiv.org/abs/2506.18588
PDF
-
Quantum-Classical Hybrid Quantized Neural Network
2025-06-23 02:12:36Wenxin Li, Chuan Wang, Hongdong Zhu, Qi Gao, Yin Ma, Hai Wei, Kai WenAbstract
Here in this work, we present a novel Quadratic Binary Optimization (QBO) model for quantized neural network training, enabling the use of arbitrary activation and loss functions through spline interpolation. We introduce Forward Interval Propagation (FIP), a method designed to tackle the challenges of non-linearity and the multi-layer composite structure in neural networks by discretizing activation functions into linear subintervals. This approach preserves the universal approximation properties of neural networks while allowing complex nonlinear functions to be optimized using quantum computers, thus broadening their applicability in artificial intelligence. We provide theoretical upper bounds on the approximation error and the number of Ising spins required, by deriving the sample complexity of the empirical risk minimization problem, from an optimization perspective. A significant challenge in solving the associated Quadratic Constrained Binary Optimization (QCBO) model on a large scale is the presence of numerous constraints. When employing the penalty method to handle these constraints, tuning a large number of penalty coefficients becomes a critical hyperparameter optimization problem, increasing computational complexity and potentially affecting solution quality. To address this, we employ the Quantum Conditional Gradient Descent (QCGD) algorithm, which leverages quantum computing to directly solve the QCBO problem. We prove the convergence of QCGD under a quantum oracle with randomness and bounded variance in objective value, as well as under limited precision constraints in the coefficient matrix. Additionally, we provide an upper bound on the Time-To-Solution for the QCBO solving process. Experimental results using a coherent Ising machine (CIM) demonstrate a 94.95% accuracy on the Fashion MNIST classification task, with only 1.1-bit precision.
Abstract (translated)
在这项工作中，我们提出了一种新颖的二次二进制优化（QBO）模型用于量化神经网络训练，并通过样条插值实现任意激活和损失函数的应用。我们还引入了前向区间传播（FIP），这是一种旨在解决非线性和多层复合结构带来的挑战的方法，它将激活函数离散化为线性子区间。这种方法保留了神经网络的通用近似性质，同时允许使用量子计算机对复杂的非线性功能进行优化，从而扩大其在人工智能领域的适用范围。 我们从优化的角度推导出经验风险最小化问题的样本复杂度，并提供理论上的误差上限和所需的伊辛自旋数量，以此来衡量近似的精度。解决与QBO模型相关的二次约束二进制优化（QCBO）问题的一个重大挑战在于存在大量的约束条件。在使用惩罚方法处理这些限制时，调整大量惩罚系数成为一个关键的超参数优化问题，并可能增加计算复杂性和影响解决方案质量。 为了解决这个问题，我们采用了量子条件梯度下降（QCGD）算法，该算法利用量子计算机直接解决QCBO问题。我们在随机量子Oracle和目标值有界方差以及系数矩阵精度有限的情况下证明了QCGD的收敛性，并提供了QCBO求解过程的时间复杂度上限。 实验结果表明，在使用相干伊辛机（CIM）处理Fashion MNIST分类任务时，仅采用1.1位精度就达到了94.95%的准确率。
URL
https://arxiv.org/abs/2506.18240
PDF
-
Robustness Evaluation of OCR-based Visual Document Understanding under Multi-Modal Adversarial Attacks
2025-06-19 15:38:31Dong Nguyen Tien, Dung D. LeAbstract
Visual Document Understanding (VDU) systems have achieved strong performance in information extraction by integrating textual, layout, and visual signals. However, their robustness under realistic adversarial perturbations remains insufficiently explored. We introduce the first unified framework for generating and evaluating multi-modal adversarial attacks on OCR-based VDU models. Our method covers six gradient-based layout attack scenarios, incorporating manipulations of OCR bounding boxes, pixels, and texts across both word and line granularities, with constraints on layout perturbation budget (e.g., IoU >= 0.6) to preserve plausibility. Experimental results across four datasets (FUNSD, CORD, SROIE, DocVQA) and six model families demonstrate that line-level attacks and compound perturbations (BBox + Pixel + Text) yield the most severe performance degradation. Projected Gradient Descent (PGD)-based BBox perturbations outperform random-shift baselines in all investigated models. Ablation studies further validate the impact of layout budget, text modification, and adversarial transferability.
Abstract (translated)
视觉文档理解（VDU）系统通过整合文本、布局和视觉信号，在信息提取方面取得了显著的性能。然而，这些系统的鲁棒性在面对现实中的对抗性扰动时仍不够充分研究。我们介绍了首个用于生成和评估基于OCR的VDU模型多模态对抗攻击的统一框架。我们的方法涵盖了六种基于梯度的布局攻击场景，包括对OCR边界框、像素和文本（从单词级别到行级别）的操作，并在布局干扰预算（例如IoU >= 0.6）上设置了约束条件以保持现实性。通过四个数据集（FUNSD、CORD、SROIE、DocVQA）及六种模型家族的实验结果表明，行级别的攻击和复合扰动（边界框+像素+文本）导致性能下降最为严重。基于投影梯度下降（PGD）的边界框扰动在所有研究的模型中均优于随机偏移基线。进一步的消融研究表明了布局预算、文本修改以及对抗迁移性的影响力。
URL
https://arxiv.org/abs/2506.16407
PDF
-
Rethinking LLM Training through Information Geometry and Quantum Metrics
2025-06-18 19:17:47Riccardo Di SipioAbstract
Optimization in large language models (LLMs) unfolds over high-dimensional parameter spaces with non-Euclidean structure. Information geometry frames this landscape using the Fisher information metric, enabling more principled learning via natural gradient descent. Though often impractical, this geometric lens clarifies phenomena such as sharp minima, generalization, and observed scaling laws. We argue that curvature-aware approaches deepen our understanding of LLM training. Finally, we speculate on quantum analogies based on the Fubini-Study metric and Quantum Fisher Information, hinting at efficient optimization in quantum-enhanced systems.
Abstract (translated)
在大型语言模型（LLMs）中的优化过程发生在具有非欧几里得结构的高维参数空间中。信息几何学利用费雪信息度量来描绘这一景观，从而使通过自然梯度下降进行更原理性的学习成为可能。尽管这种方法通常难以实现，但它有助于阐明诸如尖锐极小值、泛化和观察到的扩展规律等现象。我们认为，考虑到曲率的方法能加深我们对LLMs训练的理解。最后，基于富比尼-施图迪度量（Fubini-Study metric）和量子费雪信息（Quantum Fisher Information），我们推测出在量子增强系统中的有效优化可能存在的类比关系。
URL
https://arxiv.org/abs/2506.15830
PDF
-
Ken Utilization Layer: Hebbian Replay Within a Student's Ken for Adaptive Knowledge Tracing
2025-06-18 00:06:28Grey Kuling, Marinka ZitnikAbstract
We introduce KUL-KT, a biologically inspired architecture for knowledge tracing (KT), combining Hebbian memory encoding with gradient-based consolidation in a scalable, input-agnostic framework. KUL-KT adapts the principle of memory consolidation in neural systems, to student modeling by introducing two key innovations: (i) a time-decaying Hebbian memory update that enables graceful forgetting, and (ii) a novel Loss-aligned Internal Target (LIT) method to compute an ideal internal state, allowing continual learning without backpropagation through time. The architecture consists of a fast Hebbian memory that captures each learner interaction via a single associative update, and a slower linear network that consolidates recalled samples through gradient descent. This design enables few-shot personalization and natural forgetting without storing raw data or relying on large cohort training. Operating entirely in embedding space, KUL-KT supports both structured (tabular) and unstructured (short-answer) inputs. Empirically, KUL-KT outperforms strong baselines on ten public KT benchmarks in rank-sensitive metrics such as nDCG and Recall@10. In a classroom deployment, KUL-KT personalized quizzes from short-answer data, leading to improved learner-perceived helpfulness and reduced difficulty (p < 0.05). Ablation studies confirm that Hebbian decay and LIT are critical for continual adaptation. Compared to a strong graph-based KT model, KUL-KT trains 1.75x faster and uses 99.01\% less memory. These results position KUL-KT as a biologically grounded, memory-efficient, and input-flexible framework for personalized learning at scale.
Abstract (translated)
我们介绍了KUL-KT，这是一种生物启发的知识追踪（KT）架构，它结合了海布记忆编码与基于梯度的巩固机制，并在一个可扩展且不受输入限制的框架中实现了这一点。KUL-KT通过引入两个关键创新，将神经系统的记忆巩固原理应用到学生建模上：(i) 一种随时间衰减的海布记忆更新方式，使优雅遗忘成为可能；(ii) 一种新颖的目标与损失对齐内部目标（LIT）方法来计算理想内部状态，从而允许不通过时间反向传播进行持续学习。该架构包括一个快速的海布记忆，每次捕捉到每个学习者互动时都会通过单一联想更新来进行记录，以及一个较慢的线性网络，它通过梯度下降巩固被召回的数据样本。这种设计使得少量个性化和自然遗忘成为可能，同时不需要存储原始数据或依赖大规模群体训练。整个操作完全在嵌入空间中进行，KUL-KT支持结构化（表格）和非结构化（简答题）输入。 实验上，在十个公开的知识追踪基准测试中，KUL-KT在nDCG和Recall@10等敏感排名的度量标准上超过了强大的基线模型。在一个教室部署案例中，KUL-KT利用简答题数据为每位学生个性化测验，提高了学习者感知到的帮助效果，并降低了难度（p < 0.05）。删减研究证实了海布衰减和LIT对于持续适应性至关重要。与一个强大的基于图的KT模型相比，KUL-KT训练速度快1.75倍且内存使用量减少了99.01%。 这些结果表明，KUL-KT作为一种生物基础、记忆效率高并且输入灵活的知识追踪框架，在大规模个性化学习中具有重要位置。
URL
https://arxiv.org/abs/2507.00032
PDF
-
Efficient and Real-Time Motion Planning for Robotics Using Projection-Based Optimization
2025-06-17 17:56:42Xuemin Chi, Hakan Girgin, Tobias L\"ow, Yangyang Xie, Teng Xue, Jihao Huang, Cheng Hu, Zhitao Liu, Sylvain CalinonAbstract
Generating motions for robots interacting with objects of various shapes is a complex challenge, further complicated by the robot geometry and multiple desired behaviors. While current robot programming tools (such as inverse kinematics, collision avoidance, and manipulation planning) often treat these problems as constrained optimization, many existing solvers focus on specific problem domains or do not exploit geometric constraints effectively. We propose an efficient first-order method, Augmented Lagrangian Spectral Projected Gradient Descent (ALSPG), which leverages geometric projections via Euclidean projections, Minkowski sums, and basis functions. We show that by using geometric constraints rather than full constraints and gradients, ALSPG significantly improves real-time performance. Compared to second-order methods like iLQR, ALSPG remains competitive in the unconstrained case. We validate our method through toy examples and extensive simulations, and demonstrate its effectiveness on a 7-axis Franka robot, a 6-axis P-Rob robot and a 1:10 scale car in real-world experiments. Source codes, experimental data and videos are available on the project webpage: this https URL
Abstract (translated)
生成机器人与各种形状物体交互的运动是一个复杂的问题，这个问题进一步因机器人的几何结构和多种期望行为而变得更加困难。当前的机器人编程工具（如逆向动力学、碰撞避免和操作规划）通常将这些问题视为约束优化问题，但许多现有的求解器往往专注于特定领域或无法有效利用几何约束条件。我们提出了一种高效的一阶方法——增广拉格朗日谱投影梯度下降法（ALSPG），该方法通过欧几里得投影、闵可夫斯基和以及基函数来利用几何投影。 研究表明，通过使用几何约束而非全部约束和梯度，ALSPG在实时性能上有了显著的提高。与二次优化方法（如iLQR）相比，在无约束情况下，ALSPG依然具有竞争力。我们通过玩具示例、广泛的仿真验证了我们的方法，并且通过实际实验展示了其在7轴Franka机器人、6轴P-Rob机器人以及1:10比例模型汽车上的有效性。 源代码、实验数据和视频可在项目网页上获得：[此链接](this https URL)。
URL
https://arxiv.org/abs/2506.14865
PDF
-
A Study of Hybrid and Evolutionary Metaheuristics for Single Hidden Layer Feedforward Neural Network Architecture
2025-06-17 04:12:58Gautam Siddharth Kashyap, Md Tabrez Nafis, Samar WazirAbstract
Training Artificial Neural Networks (ANNs) with Stochastic Gradient Descent (SGD) frequently encounters difficulties, including substantial computing expense and the risk of converging to local optima, attributable to its dependence on partial weight gradients. Therefore, this work investigates Particle Swarm Optimization (PSO) and Genetic Algorithms (GAs) - two population-based Metaheuristic Optimizers (MHOs) - as alternatives to SGD to mitigate these constraints. A hybrid PSO-SGD strategy is developed to improve local search efficiency. The findings indicate that the hybrid PSO-SGD technique decreases the median training MSE by 90 to 95 percent relative to conventional GA and PSO across various network sizes (e.g., from around 0.02 to approximately 0.001 in the Sphere function). RMHC attains substantial enhancements, reducing MSE by roughly 85 to 90 percent compared to GA. Simultaneously, RS consistently exhibits errors exceeding 0.3, signifying subpar performance. These findings underscore that hybrid and evolutionary procedures significantly improve training efficiency and accuracy compared to conventional optimization methods and imply that the Building Block Hypothesis (BBH) may still be valid, indicating that advantageous weight structures are retained during evolutionary search.
Abstract (translated)
使用随机梯度下降（SGD）训练人工神经网络（ANNs）经常遇到诸如计算成本高昂和可能陷入局部最优解等问题，这些问题归因于其对权重偏导数的依赖。因此，本研究探讨了粒子群优化（PSO）和遗传算法（GAs）——两种基于种群的元启发式优化器（MHOs），作为SGD的替代方案以缓解这些限制。开发了一种混合PSO-SGD策略来提高局部搜索效率。研究表明，与传统的GA和PSO相比，该混合PSO-SGD技术在各种网络规模下将中位训练均方误差（MSE）降低了90%到95%，例如，在Sphere函数中的误差从大约0.02降至约0.001。 RMHC（随机重启Hill-Climbing算法）也取得了显著改进，与GA相比，它使MSE减少了约85%至90%。同时，RS（随机搜索）始终表现出超过0.3的错误率，表明其性能不佳。这些发现强调了混合和进化方法在训练效率和准确性方面相较于传统优化方法有明显改善，并暗示构建块假设（BBH）可能仍然有效，这意味着有利的权重结构在进化搜索中得以保留。
URL
https://arxiv.org/abs/2506.15737
PDF
-
Attention-based Adversarial Robust Distillation in Radio Signal Classifications for Low-Power IoT Devices
2025-06-13 15:39:01Lu Zhang, Sangarapillai Lambotharan, Gan Zheng, Guisheng Liao, Basil AsSadhan, Fabio RoliAbstract
Due to great success of transformers in many applications such as natural language processing and computer vision, transformers have been successfully applied in automatic modulation classification. We have shown that transformer-based radio signal classification is vulnerable to imperceptible and carefully crafted attacks called adversarial examples. Therefore, we propose a defense system against adversarial examples in transformer-based modulation classifications. Considering the need for computationally efficient architecture particularly for Internet of Things (IoT)-based applications or operation of devices in environment where power supply is limited, we propose a compact transformer for modulation classification. The advantages of robust training such as adversarial training in transformers may not be attainable in compact transformers. By demonstrating this, we propose a novel compact transformer that can enhance robustness in the presence of adversarial attacks. The new method is aimed at transferring the adversarial attention map from the robustly trained large transformer to a compact transformer. The proposed method outperforms the state-of-the-art techniques for the considered white-box scenarios including fast gradient method and projected gradient descent attacks. We have provided reasoning of the underlying working mechanisms and investigated the transferability of the adversarial examples between different architectures. The proposed method has the potential to protect the transformer from the transferability of adversarial examples.
Abstract (translated)
由于Transformer在自然语言处理和计算机视觉等众多应用中的巨大成功，它们已经被成功应用于自动调制分类中。我们已经证明了基于Transformer的无线电信号分类对精心设计且难以察觉的对抗性样本攻击非常脆弱。因此，我们提出了一种针对基于Transformer的调制分类中对抗性示例的防御系统。考虑到计算效率架构的需求，特别是在物联网（IoT）应用或在电力供应有限的环境中操作设备的情况下，我们提出了一个用于调制分类的小型化变压器。然而，在小型化变压器中可能无法实现增强鲁棒性的方法，例如基于Transformer中的对抗训练。通过证明这一点，我们提出了一种新的紧凑型变压器，可以在存在对抗性攻击时提高其健壮性。新方法旨在将来自稳健大型Transformer的对抗注意力图转移到紧凑型Transformer中。在考虑的白盒场景（包括快速梯度法和投影梯度下降攻击）下，所提方法优于现有的最先进技术。我们提供了该方法背后的工作机制的理由，并探讨了不同架构之间对抗样本的可迁移性。所提出的方法有望保护变压器免受对抗样本传输的影响。
URL
https://arxiv.org/abs/2506.11892
PDF
-
Vision Transformer with Adversarial Indicator Token against Adversarial Attacks in Radio Signal Classifications
2025-06-13 15:21:54Lu Zhang, Sangarapillai Lambotharan, Gan Zheng, Guisheng Liao, Xuekang Liu, Fabio Roli, Carsten MapleAbstract
The remarkable success of transformers across various fields such as natural language processing and computer vision has paved the way for their applications in automatic modulation classification, a critical component in the communication systems of Internet of Things (IoT) devices. However, it has been observed that transformer-based classification of radio signals is susceptible to subtle yet sophisticated adversarial attacks. To address this issue, we have developed a defensive strategy for transformer-based modulation classification systems to counter such adversarial attacks. In this paper, we propose a novel vision transformer (ViT) architecture by introducing a new concept known as adversarial indicator (AdvI) token to detect adversarial attacks. To the best of our knowledge, this is the first work to propose an AdvI token in ViT to defend against adversarial attacks. Integrating an adversarial training method with a detection mechanism using AdvI token, we combine a training time defense and running time defense in a unified neural network model, which reduces architectural complexity of the system compared to detecting adversarial perturbations using separate models. We investigate into the operational principles of our method by examining the attention mechanism. We show the proposed AdvI token acts as a crucial element within the ViT, influencing attention weights and thereby highlighting regions or features in the input data that are potentially suspicious or anomalous. Through experimental results, we demonstrate that our approach surpasses several competitive methods in handling white-box attack scenarios, including those utilizing the fast gradient method, projected gradient descent attacks and basic iterative method.
Abstract (translated)
变压器在自然语言处理和计算机视觉等多个领域的显著成功，为物联网设备通信系统中的自动调制分类应用铺平了道路。然而，观察发现基于变压器的无线电信号分类容易受到细微而复杂的对抗性攻击的影响。为了应对这一问题，我们开发了一种防御策略，用于保护基于变压器的调制分类系统的免受此类对抗性攻击。在本文中，我们提出了一种新的视觉变压器（ViT）架构，通过引入一种名为“对抗指示符”(AdvI)标记的新概念来检测对抗性攻击。据我们所知，这是首次提议使用带有AdvI标记的ViT来防御对抗性攻击。 我们将对抗训练方法与基于AdvI标记的检测机制相结合，在统一的神经网络模型中同时实现了训练时间防护和运行时防护，从而简化了系统的架构复杂度，相比单独使用的检测模型而言更为简单。通过研究注意力机制的操作原理，我们展示了提出的AdvI标记在ViT中的关键作用：它影响注意权重，并且突出显示输入数据中可能可疑或异常的区域或特征。 实验结果显示，我们的方法在处理包括快速梯度法、投影梯度下降攻击和基本迭代法在内的白盒攻击场景方面超越了多种竞争性方法。
URL
https://arxiv.org/abs/2507.00015
PDF
-
Machine Learning-Based Quantification of Vesicoureteral Reflux with Enhancing Accuracy and Efficiency
2025-06-13 07:09:12Muhyeeddin Alqaraleh, Mowafaq Salem Alzboon, Mohammad Subhi Al-Batah, Lana Yasin Al Aesa, Mohammed Hasan Abu-Arqoub, Rashiq Rafiq Marie, Firas Hussein AlsmadAbstract
Vesicoureteral reflux (VUR) is traditionally assessed using subjective grading systems, which introduces variability in diagnosis. This study investigates the use of machine learning to improve diagnostic consistency by analyzing voiding cystourethrogram (VCUG) images. A total of 113 VCUG images were reviewed, with expert grading of VUR severity. Nine image-based features were selected to train six predictive models: Logistic Regression, Decision Tree, Gradient Boosting, Neural Network, and Stochastic Gradient Descent. The models were evaluated using leave-one-out cross-validation. Analysis identified deformation patterns in the renal calyces as key indicators of high-grade VUR. All models achieved accurate classifications with no false positives or negatives. High sensitivity to subtle image patterns characteristic of different VUR grades was confirmed by substantial Area Under the Curve (AUC) values. The results suggest that machine learning can offer an objective and standardized alternative to current subjective VUR assessments. These findings highlight renal calyceal deformation as a strong predictor of severe cases. Future research should aim to expand the dataset, refine imaging features, and improve model generalizability for broader clinical use.
Abstract (translated)
输尿管膀胱反流（Vesicoureteral reflux，简称 VUR）的传统评估方法依赖于主观分级系统，这会导致诊断的一致性较差。本研究探讨了利用机器学习技术分析排尿膀胱造影（Voiding Cystourethrogram，VCUG）图像来提高诊断一致性的可能性。研究团队审查了总共113幅 VCUG 图像，并由专家对其 VUR 的严重程度进行了分级。随后选择了九个基于图像的特征，用于训练六种预测模型：逻辑回归、决策树、梯度提升、神经网络和随机梯度下降。通过留一交叉验证（leave-one-out cross-validation）方法评估这些模型的效果。 分析发现，在肾盏中观察到的变形模式是高级别 VUR 的关键指标之一。所有模型均能够准确分类，且无误报或漏报情况发生。通过对不同级别的VUR特有的细微图像特征具有高度敏感性的确认（通过显著的曲线下面积值即AUC值得出），证实了机器学习方法在识别这些特征上的有效性。 研究结果表明，机器学习可以为当前主观评估 VUR 的方式提供一个客观且标准化的选择方案。此外，该研究表明肾盏变形是严重病例的重要预测因素。未来的研究应该致力于扩大数据集规模、优化成像特征以及改进模型的泛化能力，从而使其在更广泛的临床应用中更具实用性。
URL
https://arxiv.org/abs/2506.11508
PDF
-
Generalization or Hallucination? Understanding Out-of-Context Reasoning in Transformers
2025-06-12 16:50:45Yixiao Huang, Hanlin Zhu, Tianyu Guo, Jiantao Jiao, Somayeh Sojoudi, Michael I. Jordan, Stuart Russell, Song MeiAbstract
Large language models (LLMs) can acquire new knowledge through fine-tuning, but this process exhibits a puzzling duality: models can generalize remarkably from new facts, yet are also prone to hallucinating incorrect information. However, the reasons for this phenomenon remain poorly understood. In this work, we argue that both behaviors stem from a single mechanism known as out-of-context reasoning (OCR): the ability to deduce implications by associating concepts, even those without a causal link. Our experiments across five prominent LLMs confirm that OCR indeed drives both generalization and hallucination, depending on whether the associated concepts are causally related. To build a rigorous theoretical understanding of this phenomenon, we then formalize OCR as a synthetic factual recall task. We empirically show that a one-layer single-head attention-only transformer with factorized output and value matrices can learn to solve this task, while a model with combined weights cannot, highlighting the crucial role of matrix factorization. Our theoretical analysis shows that the OCR capability can be attributed to the implicit bias of gradient descent, which favors solutions that minimize the nuclear norm of the combined output-value matrix. This mathematical structure explains why the model learns to associate facts and implications with high sample efficiency, regardless of whether the correlation is causal or merely spurious. Ultimately, our work provides a theoretical foundation for understanding the OCR phenomenon, offering a new lens for analyzing and mitigating undesirable behaviors from knowledge injection.
Abstract (translated)
大型语言模型（LLMs）可以通过微调获得新知识，但这一过程表现出一种令人费解的二元性：这些模型可以从新的事实中进行显著地概括推理，但也容易产生不正确的信息。然而，导致这种现象的原因尚不清楚。在本项工作中，我们提出两种行为都源于单一机制，即跨背景推理（OCR）：即使没有因果关系，也能够通过关联概念推断出结论的能力。我们的实验涵盖了五个知名的大规模语言模型，并确认了 OCR 确实驱动着概括和幻觉的产生，这取决于相关联的概念是否具有因果联系。 为了建立这一现象的严谨理论理解，我们随后将 OCR 形式化为一个合成的事实回忆任务。通过实验，我们证明了一层单一头部注意力机制且输出和值矩阵因子分解的变压器能够学会解决该任务，而权重合并的模型则不能，这突显了矩阵分解的关键作用。 我们的理论分析表明，OCR 能力可以归因于梯度下降的隐含偏置，这种偏置倾向于选择使组合输出-值矩阵的核范数最小化的解决方案。这一数学结构解释了为什么模型能够以高样本效率学会将事实与结论关联起来，无论这些关联是因果关系还是仅仅偶然相关。 最终，我们的工作为理解 OCR 现象提供了理论基础，并提供了一种新的视角来分析和缓解知识注入带来的不良行为。
URL
https://arxiv.org/abs/2506.10887
PDF
