arXiv:2607. 05201v1 Announce Type: new Abstract: In non-stationary streaming environments, simultaneously adapting to complex, non-linear domain shifts via continual learning while mitigating the catastrophic effects of severe, uncalibrated label noise poses a fundamental mathematical challenge.
arXiv:2607. 09785v1 Announce Type: cross Abstract: Traditionally, continual learning has assumed access to labeled data, yet many real-world applications -- such as lifelong robotics -- require models to adapt continuously from unlabeled streams.
The paper proposes Self‑Distillation Fine‑Tuning (SDFT) as a method to recover performance in Large Language Models that has been degraded by catastrophic forgetting, quantization, or pruning. It shows that SDFT restores model capabilities by aligning the high‑dimensional manifold of the student model’s hidden layers with that of a teacher model, as measured by Centered Kernel Alignment (CKA). The authors provide both empirical evidence of strong correlation between manifold alignment and performance recovery and a theoretical explanation linking generative capability to the structure of these manifolds.
The paper introduces TMLN (Trajectory-Modulatory Landscape Navigation), a method that treats continual learning as an optimal control problem on a curved loss landscape. It uses a diagonal empirical Fisher Information Matrix to approximate a local Riemannian manifold and dynamically modulates a preconditioner based on the network’s historical parameter trajectory. This trajectory‑based preconditioning is integrated into gradient updates to protect important parameter directions without adding explicit penalties, and experiments on class‑ and domain‑incremental benchmarks show a significant reduction in the loss barrier between tasks.
Point cloud denoising is essentially a geometric recovery task that aims to reconstruct the intrinsic structure of a smooth 2D Riemannian manifold embedded in R^3 from noisy, discrete ambient-space samples. Despite the remarkable progress of modern manifold-aware encoders and generative transport models in geometric representation learning, a fundamental objective-geometry mismatch remains underexplored.
arXiv:2608.21487v1 Announce Type: cross Abstract: Vision-Language Models (VLMs) exhibit strong zero-shot capabilities, making them an attractive solution for continual learning across diverse tasks....
arXiv:2606. 07086v1 Announce Type: cross Abstract: Deep neural networks (DNNs) excel in computer vision tasks given large annotated datasets.
arXiv:2607. 07513v1 Announce Type: new Abstract: Self-supervised learning matches supervised accuracy from a fraction of the labels, but the labeled-sample efficiency behind this has lacked a theoretical explanation.
The paper investigates diffusion models trained in a lazy high‑dimensional regime, extending benign overfitting theory to generative settings. By analyzing denoising score matching in a vector‑valued RKHS with an inner‑product kernel, the authors derive exact risk trajectories under gradient flow when the number of samples scales proportionally with dimensionality. These trajectories reveal three distinct phases—spectral generalization, noise‑dominated interpolation, and empirical Bayes memorization—whose interplay shapes the distribution of generated samples.
arXiv:2205. 07739v4 Announce Type: replace-cross Abstract: Self-training (ST) is a simple yet effective semi-supervised learning method.
arXiv:2607. 16231v1 Announce Type: new Abstract: Modern neural networks can fit corrupted training labels, making noisy-label learning a useful setting for studying memorization-driven overfitting.
Zero-Shot Self-Supervised Learning (ZS-SSL) has emerged as a promising paradigm for accelerated Magnetic Resonance Imaging (MRI) reconstruction, eliminating the reliance on fully-sampled external datasets. However, learning solely from a single under-sampled scan suffers from supervision scarcity and optimization instability, often leading to overfitting or artifacts.
The paper introduces the Sparse Landmark Embedding (SLE) kernel, a new framework that removes the need for conditionally negative definite (CND) distance measures in kernel methods and Gaussian Processes. By embedding each input into a sparse feature vector using compactly supported bump functions centered at all training points, any standard positive semi-definite (PSD) kernel can be applied in this embedding space, guaranteeing PSD for arbitrary distance measures. The authors provide theoretical guarantees on PSD, sparsity, stability, and universal approximation, and show through experiments with geodesic and Wasserstein distances that the SLE kernel matches or surpasses domain-specific baselines in predictive accuracy and uncertainty quantification.