arXiv subjects

Makoto Yamada

Publications and source records attributed to Makoto Yamada.

At least 19 recordsLinked to original sources

arXiv2026

SiamJEPA: On the Role of Siamese Student Encoders in JEPA

Joint Embedding Predictive Architectures (JEPAs) have emerged as a promising framework for self-supervised representation learning by predicting latent embeddings of masked regions rather than reconstructing pixels. Existing JEPA methods typically employ a single student encoder, leaving the role of Siamese student encoders largely unexplored. In this paper, we propose Siamese JEPA (SiamJEPA), a JEPA framework with masked Siamese student encoders and an exponential moving average (EMA) teacher, which can also be viewed as a JEPA formulation of the brain-inspired representation learning model PhiNet. We further introduce Random Shuffle Teacher (RST), which removes spatial correspondence in teacher targets to encourage semantic patch representations, and develop an RST-based semantic-to-spatial curriculum. Experiments on ImageNet show that Siamese student encoders effectively regularize the JEPA objective, improving representation separability and accelerating early-stage learning. Moreover, under RST, stronger Siamese regularization substantially increases class-discriminative information in individual patch tokens, suggesting that the semantic bias arises from the interaction between RST and the Siamese objective rather than from shuffling alone. SiamJEPA consistently outperforms comparable single-encoder JEPA variants under limited training budgets. With RST-based curriculum learning and a ViT-Base backbone, SiamJEPA achieves 74.2\% linear-probing accuracy after 450 epochs using a substantially simpler masking strategy, compared with I-JEPA (72.9\%) and DSeq-JEPA (73.5\%) after 600 epochs. These results demonstrate that Siamese student encoders provide an effective inductive bias for predictive representation learning and can be further enhanced through semantic-to-spatial curriculum learning. The source code is publicly available at https://github.com/oist/SiamJEPA.

cs.CV↗

arXiv2026

Neither Adversarial Training Nor Purification: Emergent Adversarial Robustness from Oscillatory Predictive Learning

Adversarial robustness in computer vision is still largely achieved through adversarial training or test-time adversarial purification, both of which introduce significant computational overhead by generating adversarial examples during training or performing iterative denoising at test time. We study whether empirical robustness can instead emerge from architectural and representation-learning inductive biases. We introduce Oscillatory Predictive Learning (OPL), a two-stage framework that combines Artificial Kuramoto Oscillatory Neurons (AKOrN) with predictive self-supervised pretraining using X-PhiNet. Because our default checkpoint uses randomized initial oscillator states, we compare it with other randomized adversarial defense methods that provide precise, reproducible, and strong attack protocols. Experiments on CIFAR-10 and CIFAR-100, with additional corruption evaluation on CIFAR-10-C, demonstrate that our method achieves competitive results under the AutoAttack-rand evaluation protocol. On CIFAR-10 and CIFAR-100, OPL attains 76.63$\pm$0.76$\%$ and 50.44$\%$ robust accuracy, respectively, under $\ell_\infty$, $ε=8/255$, AutoAttack-rand with EoT $K=20$.

cs.LG↗

arXiv2026

Crane: Context-Guided Prompt Learning and Attention Refinement for Zero-Shot Anomaly Detection

Zero-shot anomaly detection and localization aims to learn from source-domain data and generalize to unseen target domains without target-domain samples. Recent CLIP-based methods perform inference by comparing visual features with normal and abnormal textual prototypes; however, dense localization remains substantially weaker than image-level detection, especially for small or subtle abnormal regions. We identify two key limitations behind this gap: CLIP's vision encoder is primarily optimized for global image-text alignment, which limits its ability to preserve fine-grained spatial details, and the learned normal/abnormal text representations are not sufficiently aligned with dense visual features. To address these limitations, we propose Crane (Context-Guided Prompt Learning and Attention Refinement), a CLIP-based framework that first adapts the vision encoder with a correlation-based attention module to better preserve local visual structure. Second, it conditions learnable normal and abnormal prompts on global image context during training, improving instance-aware text-vision alignment. Third, it incorporates anomaly-aware local-to-global fusion, which injects anomaly-relevant patch features into the global image representation for more sensitive image-level detection. Finally, we show that the same correlation-based attention design can exploit spatial correlations from DINOv2, yielding a boosted variant, Crane+, with stronger localization ability. Across seven industrial benchmarks, Crane improves mean image-level AP by 4.5% over the strongest compared baseline, while Crane+ improves mean pixel-level AUPRO by 9.0%. The code is available at https://github.com/Alireza99Salehi/Crane.

cs.CV↗

arXiv2026

Effects of width-dependent model hyperparameters and $\ell_2$-regularization on the loss landscape of two-layer ReLU networks

Understanding deep neural networks remains a central challenge in machine learning. In particular, the theoretical properties of even two-layer ReLU networks, especially in the presence of weight decay, remain poorly understood. To this end, we derive a sufficient condition on the hyperparameter settings under which the global minima collapse to the zero solution. Interestingly, our experiments reveal that using AdamW as an optimizer prevents the collapse of the learned parameters, whereas using SGD does not, which may help explain the success of AdamW in deep learning training. In addition, when restricting the input dimension to one, we derive an analytical solution for the globally optimal parameter sets of two-layer ReLU networks and show that $\ell_2$-regularization has a width-invariant effect on connectivity, but its dimensionality-reducing effect becomes stronger as the network width increases. These results provide insight into how width-dependent hyperparameters influence the geometry of regularized loss landscapes.

cs.LG↗

arXiv2026

Task Vector Bases: A Unified and Scalable Framework for Compressed Task Arithmetic

Task arithmetic, representing downstream tasks through linear operations on task vectors, has emerged as a simple yet powerful paradigm for transferring knowledge across diverse settings. However, maintaining a large collection of task vectors introduces scalability challenges in both storage and computation. We propose Task Vector Bases, a framework compressing $T$ task vectors into $M < T$ basis vectors while preserving the functionality of task arithmetic. By representing each task vector as a structured linear combination of basis atoms, our approach supports standard operations such as addition, negation, as well as more advanced arithmetic ones. The framework is orthogonal to other efficiency-oriented improvements in task arithmetic and can be used in combination with them. We provide theoretical analysis showing that basis compression retains addition generalization guarantees and enables principled unlearning, with error bounds depending on reconstruction quality. Empirically, our proposed basis construction methods consistently outperform heuristic basis construction baselines and, in some cases, even surpass the performance of full task vector collections across diverse downstream applications while reducing storage and computational requirements. The code is available at https://github.com/uiuctml/TaskVectorBasis.

cs.LG↗

arXiv2026

Brain-Inspired Stochastic Joint Embedding Representation Learning

Representation learning is one of the key research topics in machine learning, and the framework of self-supervised learning (SSL) has revolutionized computer vision. However, these approaches have not yet fully leveraged insights from biological visual processing systems. In this paper, we introduce PhiNet v2, a novel architecture that processes temporal visual input (i.e., sequences of images) without relying on strong data augmentation, enabling it to learn robust visual representations in a manner similar to human visual processing. Our learning objective is derived from variational inference. Through extensive experiments, we demonstrate that PhiNet v2 achieves competitive performance compared to state-of-the-art vision representation models, including RSP and CropMAE, while retaining the ability to learn effectively from sequential input without strong data augmentation. This work represents a step toward more biologically plausible computer vision systems that process visual information in a manner more aligned with human cognitive processes.

cs.CV↗

arXiv2026

Delayed Momentum Aggregation: Communication-efficient Byzantine-robust Federated Learning with Partial Participation

Partial participation is essential for communication-efficient federated learning at scale, yet existing Byzantine-robust methods typically assume full client participation. In the partial participation setting, a majority of the sampled clients may be Byzantine, once Byzantine clients dominate, existing methods break down immediately. We introduce delayed momentum aggregation, a principle where the central server aggregates cached momentum from non-sampled clients along with fresh momentum from sampled clients. This principle ensures Byzantine clients remain a minority from the server's perspective even when they dominate the sampled set. We instantiate this principle in our optimizer DeMoA. We analyze the convergence rate of DeMoA, showing that DeMoA is Byzantine-robust under partial participation. Experiments show that, with 20% Byzantine ratio and only 10% partial participation rate, DeMoA achieves the best accuracy even when existing methods fail empirically.

cs.LG↗

arXiv2026

On Verbalized Confidence Scores for LLMs

The rise of large language models (LLMs) and their tight integration into our daily life make it essential to dedicate efforts towards their trustworthiness. Uncertainty quantification for LLMs can establish more human trust into their responses, but also allows LLM agents to make more informed decisions based on each other's uncertainty. To estimate the uncertainty in a response, internal token logits, task-specific proxy models, or sampling of multiple responses are commonly used. This work focuses on asking the LLM itself to verbalize its uncertainty with a confidence score as part of its output tokens, which is a promising way for prompt- and model-agnostic uncertainty quantification with low overhead. Using an extensive benchmark, we assess the reliability of verbalized confidence scores with respect to different datasets, models, and prompt methods. Our results reveal that the reliability of these scores strongly depends on how the model is asked, but also that it is possible to extract well-calibrated confidence scores with certain prompt methods. We argue that verbalized confidence scores can become a simple but effective and versatile uncertainty quantification method in the future. Our code is available at https://github.com/danielyxyang/llm-verbalized-uq.

cs.CL↗

arXiv2026

TIPS Over Tricks: Simple Prompts for Effective Zero-shot Anomaly Detection

Anomaly detection identifies departures from expected behavior in safety-critical settings. When target-domain normal data are unavailable, zero-shot anomaly detection (ZSAD) leverages vision-language models (VLMs). However, CLIP's coarse image-text alignment limits both localization and detection due to (i) spatial misalignment and (ii) weak sensitivity to fine-grained anomalies; prior works compensate with complex auxiliary modules yet largely overlook the choice of backbone. We revisit the backbone and use TIPS-a VLM trained with spatially aware objectives. While TIPS alleviates CLIP's issues, it exposes a distributional gap between global and local features. We address this with decoupled prompts-fixed for image-level detection and learnable for pixel-level localization-and by injecting local evidence into the global score. Without CLIP-specific tricks, our TIPS-based pipeline improves image-level performance by 1.1-3.9% and pixel-level by 1.5-6.9% across seven industrial datasets, delivering strong generalization with a lean architecture. Code is available at github.com/AlirezaSalehy/Tipsomaly.

cs.CV↗

arXiv2025

When LRP Diverges from Leave-One-Out in Transformers

Leave-One-Out (LOO) provides an intuitive measure of feature importance but is computationally prohibitive. While Layer-Wise Relevance Propagation (LRP) offers a potentially efficient alternative, its axiomatic soundness in modern Transformers remains largely under-examined. In this work, we first show that the bilinear propagation rules used in recent advances of AttnLRP violate the implementation invariance axiom. We prove this analytically and confirm it empirically in linear attention layers. Second, we also revisit CP-LRP as a diagnostic baseline and find that bypassing relevance propagation through the softmax layer -- backpropagating relevance only through the value matrices -- significantly improves alignment with LOO, particularly in middle-to-late Transformer layers. Overall, our results suggest that (i) bilinear factorization sensitivity and (ii) softmax propagation error potentially jointly undermine LRP's ability to approximate LOO in Transformers.

cs.LG↗

arXiv2025

Sparse minimum Redundancy Maximum Relevance for feature selection

We propose a feature screening method that integrates both feature-feature and feature-target relationships. Inactive features are identified via a penalized minimum Redundancy Maximum Relevance (mRMR) procedure, which is the continuous version of the classic mRMR penalized by a non-convex regularizer, and where the parameters estimated as zero coefficients represent the set of inactive features. We establish the conditions under which zero coefficients are correctly identified to guarantee accurate recovery of inactive features. We introduce a multi-stage procedure based on the knockoff filter enabling the penalized mRMR to discard inactive features while controlling the false discovery rate (FDR). Our method performs comparably to HSIC-LASSO but is more conservative in the number of selected features. It only requires setting an FDR threshold, rather than specifying the number of features to retain. The effectiveness of the method is illustrated through simulations and real-world datasets. The code to reproduce this work is available on the following GitHub: https://github.com/PeterJackNaylor/SmRMR.

stat.ML↗

arXiv2025

Data Poisoning for In-context Learning

In the domain of large language models (LLMs), in-context learning (ICL) has been recognized for its innovative ability to adapt to new tasks, relying on examples rather than retraining or fine-tuning. This paper delves into the critical issue of ICL's susceptibility to data poisoning attacks, an area not yet fully explored. We wonder whether ICL is vulnerable, with adversaries capable of manipulating example data to degrade model performance. To address this, we introduce ICLPoison, a specialized attacking framework conceived to exploit the learning mechanisms of ICL. Our approach uniquely employs discrete text perturbations to strategically influence the hidden states of LLMs during the ICL process. We outline three representative strategies to implement attacks under our framework, each rigorously evaluated across a variety of models and tasks. Our comprehensive tests, including trials on the sophisticated GPT-4 model, demonstrate that ICL's performance is significantly compromised under our framework. These revelations indicate an urgent need for enhanced defense mechanisms to safeguard the integrity and reliability of LLMs in applications relying on in-context learning.

cs.CR↗

arXiv2025

Towards the Effect of Examples on In-Context Learning: A Theoretical Case Study

In-context learning (ICL) has emerged as a powerful capability for large language models (LLMs) to adapt to downstream tasks by leveraging a few (demonstration) examples. Despite its effectiveness, the mechanism behind ICL remains underexplored. To better understand how ICL integrates the examples with the knowledge learned by the LLM during pre-training (i.e., pre-training knowledge) and how the examples impact ICL, this paper conducts a theoretical study in binary classification tasks. In particular, we introduce a probabilistic model extending from the Gaussian mixture model to exactly quantify the impact of pre-training knowledge, label frequency, and label noise on the prediction accuracy. Based on our analysis, when the pre-training knowledge contradicts the knowledge in the examples, whether ICL prediction relies more on the pre-training knowledge or the examples depends on the number of examples. In addition, the label frequency and label noise of the examples both affect the accuracy of the ICL prediction, where the minor class has a lower accuracy, and how the label noise impacts the accuracy is determined by the specific noise level of the two classes. Extensive simulations are conducted to verify the correctness of the theoretical results, and real-data experiments also align with the theoretical insights. Our work reveals the role of pre-training knowledge and examples in ICL, offering a deeper understanding of LLMs' behaviors in classification tasks.

cs.LG↗

arXiv2025

Learning Structured Representations by Embedding Class Hierarchy with Fast Optimal Transport

To embed structured knowledge within labels into feature representations, prior work [Zeng et al., 2022] proposed to use the Cophenetic Correlation Coefficient (CPCC) as a regularizer during supervised learning. This regularizer calculates pairwise Euclidean distances of class means and aligns them with the corresponding shortest path distances derived from the label hierarchy tree. However, class means may not be good representatives of the class conditional distributions, especially when they are multi-mode in nature. To address this limitation, under the CPCC framework, we propose to use the Earth Mover's Distance (EMD) to measure the pairwise distances among classes in the feature space. We show that our exact EMD method generalizes previous work, and recovers the existing algorithm when class-conditional distributions are Gaussian. To further improve the computational efficiency of our method, we introduce the Optimal Transport-CPCC family by exploring four EMD approximation variants. Our most efficient OT-CPCC variant, the proposed Fast FlowTree algorithm, runs in linear time in the size of the dataset, while maintaining competitive performance across datasets and tasks. The code is available at https://github.com/uiuctml/OTCPCC.

cs.LG↗

arXiv2025

Many-to-Many Matching via Sparsity Controlled Optimal Transport

Many-to-many matching seeks to match multiple points in one set and multiple points in another set, which is a basis for a wide range of data mining problems. It can be naturally recast in the framework of Optimal Transport (OT). However, existing OT methods either lack the ability to accomplish many-to-many matching or necessitate careful tuning of a regularization parameter to achieve satisfactory results. This paper proposes a novel many-to-many matching method to explicitly encode many-to-many constraints while preventing the degeneration into one-to-one matching. The proposed method consists of the following two components. The first component is the matching budget constraints on each row and column of a transport plan, which specify how many points can be matched to a point at most. The second component is the deformed $q$-entropy regularization, which encourages a point to meet the matching budget maximally. While the deformed $q$-entropy was initially proposed to sparsify a transport plan, we employ it to avoid the degeneration into one-to-one matching. We optimize the objective via a penalty algorithm, which is efficient and theoretically guaranteed to converge. Experimental results on various tasks demonstrate that the proposed method achieves good performance by gleaning meaningful many-to-many matchings.

cs.LG↗

arXiv2025

PhiNets: Brain-inspired Non-contrastive Learning Based on Temporal Prediction Hypothesis

Predictive coding is a theory which hypothesises that cortex predicts sensory inputs at various levels of abstraction to minimise prediction errors. Inspired by predictive coding, Chen et al. (2024) proposed another theory, temporal prediction hypothesis, to claim that sequence memory residing in hippocampus has emerged through predicting input signals from the past sensory inputs. Specifically, they supposed that the CA3 predictor in hippocampus creates synaptic delay between input signals, which is compensated by the following CA1 predictor. Though recorded neural activities were replicated based on the temporal prediction hypothesis, its validity has not been fully explored. In this work, we aim to explore the temporal prediction hypothesis from the perspective of self-supervised learning. Specifically, we focus on non-contrastive learning, which generates two augmented views of an input image and predicts one from another. Non-contrastive learning is intimately related to the temporal prediction hypothesis because the synaptic delay is implicitly created by StopGradient. Building upon a popular non-contrastive learner, SimSiam, we propose PhiNet, an extension of SimSiam to have two predictors explicitly corresponding to the CA3 and CA1, respectively. Through studying the PhiNet model, we discover two findings. First, meaningful data representations emerge in PhiNet more stably than in SimSiam. This is initially supported by our learning dynamics analysis: PhiNet is more robust to the representational collapse. Second, PhiNet adapts more quickly to newly incoming patterns in online and continual learning scenarios. For practitioners, we additionally propose an extension called X-PhiNet integrated with a momentum encoder, excelling in continual learning. All in all, our work reveals that the temporal prediction hypothesis is a reasonable model in terms of the robustness and adaptivity.

cs.LG↗

arXiv2025

Necessary and Sufficient Watermark for Large Language Models

In recent years, large language models (LLMs) have achieved remarkable performances in various NLP tasks. They can generate texts that are indistinguishable from those written by humans. Such remarkable performance of LLMs increases their risk of being used for malicious purposes, such as generating fake news articles. Therefore, it is necessary to develop methods for distinguishing texts written by LLMs from those written by humans. Watermarking is one of the most powerful methods for achieving this. Although existing watermarking methods have successfully detected texts generated by LLMs, they significantly degrade the quality of the generated texts. In this study, we propose the Necessary and Sufficient Watermark (NS-Watermark) for inserting watermarks into generated texts without degrading the text quality. More specifically, we derive minimum constraints required to be imposed on the generated texts to distinguish whether LLMs or humans write the texts. Then, we formulate the NS-Watermark as a constrained optimization problem and propose an efficient algorithm to solve it. Through the experiments, we demonstrate that the NS-Watermark can generate more natural texts than existing watermarking methods and distinguish more accurately between texts written by LLMs and those written by humans. Especially in machine translation tasks, the NS-Watermark can outperform the existing watermarking method by up to 30 BLEU scores.

cs.CL↗

arXiv2025

Fast unsupervised ground metric learning with tree-Wasserstein distance

The performance of unsupervised methods such as clustering depends on the choice of distance metric between features, or ground metric. Commonly, ground metrics are decided with heuristics or learned via supervised algorithms. However, since many interesting datasets are unlabelled, unsupervised ground metric learning approaches have been introduced. One promising option employs Wasserstein singular vectors (WSVs), which emerge when computing optimal transport distances between features and samples simultaneously. WSVs are effective, but can be prohibitively computationally expensive in some applications: $\mathcal{O}(n^2m^2(n \log(n) + m \log(m))$ for $n$ samples and $m$ features. In this work, we propose to augment the WSV method by embedding samples and features on trees, on which we compute the tree-Wasserstein distance (TWD). We demonstrate theoretically and empirically that the algorithm converges to a better approximation of the standard WSV approach than the best known alternatives, and does so with $\mathcal{O}(n^3+m^3+mn)$ complexity. In addition, we prove that the initial tree structure can be chosen flexibly, since tree geometry does not constrain the richness of the approximation up to the number of edge weights. This proof suggests a fast and recursive algorithm for computing the tree parameter basis set, which we find crucial to realising the efficiency gains at scale. Finally, we employ the tree-WSV algorithm to several single-cell RNA sequencing genomics datasets, demonstrating its scalability and utility for unsupervised cell-type clustering problems. These results poise unsupervised ground metric learning with TWD as a low-rank approximation of WSV with the potential for widespread application.

cs.LG↗