On the Learning Dynamics of Two-layer Linear Networks with Label Noise SGD

One crucial factor behind the success of deep learning lies in the implicit bias induced by noise inherent in gradient-based training algorithms. Motivated by empirical observations that training with noisy labels improves model generalization, we delve into the underlying mechanisms behind stochastic gradient descent (SGD) with label noise. Focusing on a two-layer over-parameterized linear network, we analyze the learning dynamics of label noise SGD, unveiling a two-phase learning behavior. In \emph{Phase I}, the magnitudes of model weights progressively diminish, and the model escapes the lazy regime; enters the rich regime. In \emph{Phase II}, the alignment between model weights and the ground-truth interpolator increases, and the model eventually converges. Our analysis highlights the critical role of label noise in driving the transition from the lazy to the rich regime and minimally explains its empirical success. Furthermore, we extend these insights to Sharpness-Aware Minimization (SAM), showing that the principles governing label noise SGD also apply to broader optimization algorithms. Extensive experiments, conducted under both synthetic and real-world setups, strongly support our theory. Our code is released at https://github.com/a-usually/Label-Noise-SGD.

Wiener Chaos Expansion based Neural Operator for Singular Stochastic Partial Differential Equations

In this paper, we explore how our recently developed Wiener Chaos Expansion (WCE)-based neural operator (NO) can be applied to singular stochastic partial differential equations, e.g., the dynamic $\boldsymbolΦ^4_2$ model simulated in the recent works. Unlike the previous WCE-NO which solves SPDEs by simply inserting Wick-Hermite features into the backbone NO model, we leverage feature-wise linear modulation (FiLM) to appropriately capture the dependency between the solution of singular SPDE and its smooth remainder. The resulting WCE-FiLM-NO shows excellent performance on $\boldsymbolΦ^4_2$, as measured by relative $L_2$ loss, out-of-distribution $L_2$ loss, and autocorrelation score; all without the help of renormalisation factor. In addition, we also show the potential of simulating $\boldsymbolΦ^4_3$ data, which is more aligned with real scientific practice in statistical quantum field theory. To the best of our knowledge, this is among the first works to develop an efficient data-driven surrogate for the dynamical $\boldsymbolΦ^4_3$ model.

Expanding the Chaos: Neural Operator for Stochastic (Partial) Differential Equations

Stochastic differential equations (SDEs) and stochastic partial differential equations (SPDEs) are fundamental tools for modeling stochastic dynamics across the natural sciences and modern machine learning. Developing deep learning models for approximating their solution operators promises not only fast, practical solvers, but may also inspire models that resolve classical learning tasks from a new perspective. In this work, we build on classical Wiener chaos expansions (WCE) to design neural operator (NO) architectures for SPDEs and SDEs: we project the driving noise paths onto orthonormal Wick Hermite features and parameterize the resulting deterministic chaos coefficients with neural operators, so that full solution trajectories can be reconstructed from noise in a single forward pass. On the theoretical side, we investigate the classical WCE results for the class of multi-dimensional SDEs and semilinear SPDEs considered here by explicitly writing down the associated coupled ODE/PDE systems for their chaos coefficients, which makes the separation between stochastic forcing and deterministic dynamics fully explicit and directly motivates our model designs. On the empirical side, we validate our models on a diverse suite of problems: classical SPDE benchmarks, diffusion one-step sampling on images, topological interpolation on graphs, financial extrapolation, parameter estimation, and manifold SDEs for flood prediction, demonstrating competitive accuracy and broad applicability. Overall, our results indicate that WCE-based neural operators provide a practical and scalable way to learn SDE/SPDE solution operators across diverse domains.

ACT as Human: Multimodal Large Language Model Data Annotation with Critical Thinking

Supervised learning relies on high-quality labeled data, but obtaining such data through human annotation is both expensive and time-consuming. Recent work explores using large language models (LLMs) for annotation, but LLM-generated labels still fall short of human-level quality. To address this problem, we propose the Annotation with Critical Thinking (ACT) data pipeline, where LLMs serve not only as annotators but also as judges to critically identify potential errors. Human effort is then directed towards reviewing only the most "suspicious" cases, significantly improving the human annotation efficiency. Our major contributions are as follows: (1) ACT is applicable to a wide range of domains, including natural language processing (NLP), computer vision (CV), and multimodal understanding, by leveraging multimodal-LLMs (MLLMs). (2) Through empirical studies, we derive 7 insights on how to enhance annotation quality while efficiently reducing the human cost, and then translate these findings into user-friendly guidelines. (3) We theoretically analyze how to modify the loss function so that models trained on ACT data achieve similar performance to those trained on fully human-annotated data. Our experiments show that the performance gap can be reduced to less than 2% on most benchmark datasets while saving up to 90% of human costs.

Generalized infinite dimensional Alpha-Procrustes based geometries

This work extends the recently introduced Alpha-Procrustes family of Riemannian metrics for symmetric positive definite (SPD) matrices by incorporating generalized versions of the Bures-Wasserstein (GBW), Log-Euclidean, and Wasserstein distances. While the Alpha-Procrustes framework has unified many classical metrics in both finite- and infinite- dimensional settings, it previously lacked the structural components necessary to realize these generalized forms. We introduce a formalism based on unitized Hilbert-Schmidt operators and an extended Mahalanobis norm that allows the construction of robust, infinite-dimensional generalizations of GBW and Log-Hilbert-Schmidt distances. Our approach also incorporates a learnable regularization parameter that enhances geometric stability in high-dimensional comparisons. Preliminary experiments reproducing benchmarks from the literature demonstrate the improved performance of our generalized metrics, particularly in scenarios involving comparisons between datasets of varying dimension and scale. This work lays a theoretical and computational foundation for advancing robust geometric methods in machine learning, statistical inference, and functional data analysis.

Provable Benefit of Curriculum in Transformer Tree-Reasoning Post-Training

Recent curriculum techniques in the post-training stage of LLMs have been widely observed to outperform non-curriculum approaches in enhancing reasoning performance, yet a principled understanding of why and to what extent they work remains elusive. To address this gap, we develop a theoretical framework grounded in the intuition that progressively learning through manageable steps is more efficient than directly tackling a hard reasoning task, provided each stage stays within the model's effective competence. Under mild complexity conditions linking consecutive curriculum stages, we show that curriculum post-training avoids the exponential complexity bottleneck. To substantiate this result, drawing insights from the Chain-of-Thoughts (CoTs) solving mathematical problems such as Countdown and parity, we model CoT generation as a states-conditioned autoregressive reasoning tree, define a uniform-branching base model to capture pretrained behavior, and formalize curriculum stages as either depth-increasing (longer reasoning chains) or hint-decreasing (shorter prefixes) subtasks. Our analysis shows that, under outcome-only reward signals, reinforcement learning finetuning achieves high accuracy with polynomial sample complexity, whereas direct learning suffers from an exponential bottleneck. We further establish analogous guarantees for test-time scaling, where curriculum-aware querying reduces both reward oracle calls and sampling cost from exponential to polynomial order.

Consistency Is Not Always Correct: Towards Understanding the Role of Exploration in Post-Training Reasoning

Foundation models exhibit broad knowledge but limited task-specific reasoning, motivating post-training strategies such as RLVR and inference scaling with outcome or process reward models (ORM/PRM). While recent work highlights the role of exploration and entropy stability in improving pass@K, empirical evidence points to a paradox: RLVR and ORM/PRM typically reinforce existing tree-like reasoning paths rather than expanding the reasoning scope, raising the question of why exploration helps at all if no new patterns emerge. To reconcile this paradox, we adopt the perspective of Kim et al. (2025), viewing easy (e.g., simplifying a fraction) versus hard (e.g., discovering a symmetry) reasoning steps as low- versus high-probability Markov transitions, and formalize post-training dynamics through Multi-task Tree-structured Markov Chains (TMC). In this tractable model, pretraining corresponds to tree expansion, while post-training corresponds to chain-of-thought reweighting. We show that several phenomena recently observed in empirical studies arise naturally in this setting: (1) RLVR induces a squeezing effect, reducing reasoning entropy and forgetting some correct paths; (2) population rewards of ORM/PRM encourage consistency rather than accuracy, thereby favoring common patterns; and (3) certain rare, high-uncertainty reasoning paths by the base model are responsible for solving hard problem instances. Together, these explain why exploration -- even when confined to the base model's reasoning scope -- remains essential: it preserves access to rare but crucial reasoning traces needed for difficult cases, which are squeezed out by RLVR or unfavored by inference scaling. Building on this, we further show that exploration strategies such as rejecting easy instances and KL regularization help preserve rare reasoning traces. Empirical simulations corroborate our theoretical results.

Self-Contradiction as Self-Improvement: Mitigating the Generation-Understanding Gap in MLLMs

Despite efforts to unify multimodal generation and understanding tasks in a single model, we show these MLLMs exhibit self-contradiction where generation produces images deemed misaligned with input prompts based on the model's own understanding. We define a Nonunified score that quantifies such self-contradiction. Our empirical results reveal that the self-contradiction mainly arises from weak generation that fails to align with prompts, rather than misunderstanding. This capability asymmetry indicates the potential of leveraging self-contradiction for self-improvement, where the stronger model understanding guides the weaker generation to mitigate the generation-understanding gap. Applying standard post-training methods (e.g., SFT, DPO) with such internal supervision successfully improves both generation and unification. We discover a co-improvement effect on both generation and understanding when only fine-tuning the generation branch, a phenomenon known in pre-training but underexplored in post-training. Our analysis shows improvements stem from better detection of false positives that are previously incorrectly identified as prompt-aligned. Theoretically, we show the aligned training dynamics between generation and understanding allow reduced prompt-misaligned generations to also improve mismatch detection in the understanding branch. Additionally, the framework reveals a potential risk of co-degradation under poor supervision-an overlooked phenomenon that is empirically validated in our experiments. Notably, we find intrinsic metrics like Nonunified score cannot distinguish co-degradation from co-improvement, which highlights the necessity of data quality check. Finally, we propose a curriculum-based strategy based on our findings that gradually introduces harder samples as the model improves, leading to better unification and improved MLLM generation and understanding.

Generalization Bound of Gradient Flow through Training Trajectory and Data-dependent Kernel

Gradient-based optimization methods have shown remarkable empirical success, yet their theoretical generalization properties remain only partially understood. In this paper, we establish a generalization bound for gradient flow that aligns with the classical Rademacher complexity bounds for kernel methods-specifically those based on the RKHS norm and kernel trace-through a data-dependent kernel called the loss path kernel (LPK). Unlike static kernels such as NTK, the LPK captures the entire training trajectory, adapting to both data and optimization dynamics, leading to tighter and more informative generalization guarantees. Moreover, the bound highlights how the norm of the training loss gradients along the optimization trajectory influences the final generalization performance. The key technical ingredients in our proof combine stability analysis of gradient flow with uniform convergence via Rademacher complexity. Our bound recovers existing kernel regression bounds for overparameterized neural networks and shows the feature learning capability of neural networks compared to kernel methods. Numerical experiments on real-world datasets validate that our bounds correlate well with the true generalization gap.

Scalable Parameter and Memory Efficient Pretraining for LLM: Recent Algorithmic Advances and Benchmarking

Fueled by their remarkable ability to tackle diverse tasks across multiple domains, large language models (LLMs) have grown at an unprecedented rate, with some recent models containing trillions of parameters. This growth is accompanied by substantial computational challenges, particularly regarding the memory and compute resources required for training and fine-tuning. Numerous approaches have been explored to address these issues, such as LoRA. While these methods are effective for fine-tuning, their application to pre-training is significantly more challenging due to the need to learn vast datasets. Motivated by this issue, we aim to address the following questions: Can parameter- or memory-efficient methods enhance pre-training efficiency while achieving performance comparable to full-model training? How can the performance gap be narrowed? To this end, the contributions of this work are the following. (1) We begin by conducting a comprehensive survey that summarizes state-of-the-art methods for efficient pre-training. (2) We perform a benchmark evaluation of several representative memory efficient pre-training approaches to comprehensively evaluate their performance across model sizes. We observe that with a proper choice of optimizer and hyperparameters, full-rank training delivers the best performance, as expected. We also notice that incorporating high-rank updates in low-rank approaches is the key to improving their performance. (3) Finally, we propose two practical techniques, namely weight refactorization and momentum reset, to enhance the performance of efficient pre-training methods. We observe that applying these techniques to the low-rank method (on a 1B model) can achieve a lower perplexity than popular memory efficient algorithms such as GaLore and Fira, while simultaneously using about 25% less memory.