ICML 2025 论文列表

In sparse optimization, enforcing hard constraints using the $\ell_0$ pseudo-norm offers advantages like controlled sparsity compared to convex relaxations. However, many real-world applications demand not only sparsity constraints but also some extra constraints. While prior algorithms have been developed to address this complex scenario with mixed combinatorial and convex constraints, they typically require the closed form projection onto the mixed constraints which might not exist, and/or only provide local guarantees of convergence which is different from the global guarantees commonly sought in sparse optimization. To fill this gap, in this paper, we study the problem of sparse optimization with extra *support-preserving* constraints commonly encountered in the literature. We present a new variant of iterative hard-thresholding algorithm equipped with a two-step consecutive projection operator customized for these mixed constraints, serving as a simple alternative to the Euclidean projection onto the mixed constraint. By introducing a novel trade-off between sparsity relaxation and sub-optimality, we provide global guarantees in objective value for the output of our algorithm, in the deterministic, stochastic, and zeroth-order settings, under the conventional restricted strong-convexity/smoothness assumptions. As a fundamental contribution in proof techniques, we develop a novel extension of the classic three-point lemma to the considered two-step non-convex projection operator, which allows us to analyze the convergence in objective value in an elegant way that has not been possible with existing techniques. In the zeroth-order case, such technique also improves upon the state-of-the-art result from de Vazelhes et. al. (2022), even in the case without additional constraints, by allowing us to remove a non-vanishing system error present in their work.

We study generalization of iterative noisy gradient schemes on smooth non-convex losses. Formally, we establish time-independent information theoretic generalization bounds for Stochastic Gradient Langevin Dynamics (SGLD) that do not diverge as the iteration count increases. Our bounds are obtained through a stability argument: we analyze the difference between two SGLD sequences ran in parallel on two datasets sampled from the same distribution. Our result only requires an isoperimetric inequality to hold, which is merely a restriction on the tails of the loss. Our work relaxes the assumptions of prior work to establish that the iterates stay within a bounded KL divergence from each other. Under an additional dissipativity assumption, we show that the stronger Renyi divergence also stays bounded by establishing a uniform log-Sobolev constant of the iterates. Without dissipativity, we sidestep the need for local log-Sobolev inequalities and instead exploit the regularizing properties of Gaussian convolution. These techniques allow us to show that strong convexity is not necessary for finite stability bounds. Our work shows that noisy SGD can have finite, iteration-independent, generalization and differential privacy bounds in unbounded non-convex settings.

Overparametrized Deep Neural Networks (DNNs) have demonstrated remarkable success in a wide variety of domains too high-dimensional for classical shallow networks subject to the curse of dimensionality. However, open questions about fundamental principles, that govern the learning dynamics of DNNs, remain. In this position paper we argue that it is the ability of DNNs to exploit the compositionally sparse structure of the target function driving their success. As such, DNNs can leverage the property that most practically relevant functions can be composed from a small set of constituent functions, each of which relies only on a low-dimensional subset of all inputs. We show that this property is shared by all efficiently Turing-computable functions and is therefore highly likely present in all current learning problems. While some promising theoretical insights on questions concerned with approximation and generalization exist in the setting of compositionally sparse functions, several important questions on the learnability and optimization of DNNs remain. Completing the picture of the role of compositional sparsity in deep learning is essential to a comprehensive theory of artificial—and even general—intelligence.

Despite the rapid development of large language models (LLMs), a fundamental challenge persists: the lack of high-quality optimization modeling datasets hampers LLMs' robust modeling of practical optimization problems from natural language descriptions (NL). This data scarcity also contributes to the generalization difficulties experienced by learning-based methods. To address these challenges, we propose a scalable framework for synthesizing a high-quality dataset, named OptMATH. Starting from curated seed data with mathematical formulations (MF), this framework automatically generates problem data (PD) with controllable complexity. Then, a back-translation step is employed to obtain NL. To verify the correspondence between the NL and the PD, a forward modeling step followed by rejection sampling is used. The accepted pairs constitute the training part of OptMATH. Then a collection of rejected pairs is identified and further filtered. This collection serves as a new benchmark for optimization modeling, containing difficult instances whose lengths are much longer than these of NL4OPT and MAMO. Through extensive experiments, we demonstrate that models of various sizes (0.5B-32B parameters) trained on OptMATH achieve superior results on multiple modeling benchmarks, thereby validating the effectiveness and scalability of our approach. The OptMATH dataset and related resources are available at \url{https://github.com/optsuite/OptMATH}.

Modern policy optimization methods roughly follow the policy mirror descent (PMD) algorithmic template, for which there are by now numerous theoretical convergence results. However, most of these either target tabular environments, or can be applied effectively only when the class of policies being optimized over satisfies strong closure conditions, which is typically not the case when working with parametric policy classes in large-scale environments. In this work, we develop a theoretical framework for PMD for general policy classes where we replace the closure conditions with a generally weaker variational gradient dominance assumption, and obtain upper bounds on the rate of convergence to the best-in-class policy. Our main result leverages a novel notion of smoothness with respect to a local norm induced by the occupancy measure of the current policy, and casts PMD as a particular instance of smooth non-convex optimization in non-Euclidean space.

Federated learning faces severe communication bottlenecks due to the high dimensionality of model updates. Communication compression with contractive compressors (e.g., Top-$K$) is often preferable in practice but can degrade performance without proper handling. Error feedback (EF) mitigates such issues but has been largely restricted for smooth, unconstrained problems, limiting its real-world applicability where non-smooth objectives and safety constraints are critical. We advance our understanding of EF in the canonical non-smooth convex setting by establishing new lower complexity bounds for first-order algorithms with contractive compression. Next, we propose Safe-EF, a novel algorithm that matches our lower bound (up to a constant) while enforcing safety constraints essential for practical applications. Extending our approach to the stochastic setting, we bridge the gap between theory and practical implementation. Extensive experiments in a reinforcement learning setup, simulating distributed humanoid robot training, validate the effectiveness of Safe-EF in ensuring safety and reducing communication complexity.

Recently, the expansion of Variance Reduction (VR) to Riemannian stochastic non-convex optimization has attracted increasing interest. Inspired by recursive momentum, we first introduce Stochastic Recursive Variance Reduced Gradient (SRVRG) algorithm and further present Stochastic Recursive Gradient Estimator (SRGE) in Euclidean spaces, which unifies the prevailing variance reduction estimators. We then extend SRGE to Riemannian spaces, resulting in a unified Stochastic rEcursive vaRiance reducEd gradieNt frAmework (SERENA) for Riemannian non-convex optimization. This framework includes the proposed R-SRVRG, R-SVRRM, and R-Hybrid-SGD methods, as well as other existing Riemannian VR methods. Furthermore, we establish a unified theoretical analysis for Riemannian non-convex optimization under retraction and vector transport. The IFO complexity of our proposed R-SRVRG and R-SVRRM to converge to $\varepsilon$-accurate solution is $\mathcal{O}\left(\min \{n^{1/2}{\varepsilon^{-2}}, \varepsilon^{-3}\}\right)$ in the finite-sum setting and ${\mathcal{O}\left( \varepsilon^{-3}\right)}$ for the online case, both of which align with the lower IFO complexity bound. Experimental results indicate that the proposed algorithms surpass other existing Riemannian optimization methods.

We develop a novel optimization method for quantization-aware training (QAT). Specifically, we show that convex, piecewise-affine regularization (PAR) can effectively induce neural network weights to cluster towards discrete values. We minimize PAR-regularized loss functions using an aggregate proximal stochastic gradient method (AProx) and prove that it enjoys last-iterate convergence. Our approach provides an interpretation of the straight-through estimator (STE), a widely used heuristic for QAT, as the asymptotic form of PARQ. We conduct experiments to demonstrate that PARQ obtains competitive performance on convolution- and transformer-based vision tasks.

Despite Low-Rank Adaptation (LoRA)'s popularity for fine-tuning large models, it often exhibits a noticeable performance gap compared to full fine-tuning, particularly in complex tasks such as mathematical reasoning and code generation. Motivated by this discrepancy, we propose a novel fusion approach for LoRA fine-tuned models. Our key insight is that LoRA models trained with different random seeds on the same task often exhibit complementary strengths. In contrast to existing research that typically focuses on fusing models trained on diverse tasks, we explore the potential of combining multiple LoRA models fine-tuned on the same task with different random seeds. This intra-task fusion method aims to leverage the strengths of various fine-tuned models to create a more robust and effective adaptation. To validate our approach, we conducted comprehensive experiments across three key areas: mathematical reasoning, code generation, and general instruction-tuning tasks. The results demonstrate that our fusion method significantly enhances LoRA's performance, outperforming both standalone LoRA models and current fusion methods. Notably, this advancement substantially narrows the gap between LoRA and full fine-tuning, thus offering a more effective approach to model adaptation without the GPU memory burden of full parameter fine-tuning.

The minimax sample complexity of group distributionally robust optimization (GDRO) has been determined up to a $\log(K)$ factor, where $K$ is the number of groups. In this work, we venture beyond the minimax perspective via a novel notion of sparsity that we call $(\lambda, \beta)$-sparsity. In short, this condition means that at any parameter $\theta$, there is a set of at most $\beta$ groups whose risks at $\theta$ are all at least $\lambda$ larger than the risks of the other groups. To find an $\epsilon$-optimal $\theta$, we show via a novel algorithm and analysis that the $\epsilon$-dependent term in the sample complexity can swap a linear dependence on $K$ for a linear dependence on the potentially much smaller $\beta$. This improvement leverages recent progress in sleeping bandits, showing a fundamental connection between the two-player zero-sum game optimization framework for GDRO and per-action regret bounds in sleeping bandits. We next show an adaptive algorithm which, up to logarithmic factors, obtains a sample complexity bound that adapts to the best $(\lambda, \beta)$-sparsity condition that holds. We also show how to obtain a dimension-free semi-adaptive sample complexity bound with a computationally efficient method. Finally, we demonstrate the practicality of the $(\lambda, \beta)$-sparsity condition and the improved sample efficiency of our algorithms on both synthetic and real-life datasets.

Distributed optimization has become the default training paradigm in modern machine learning due to the growing scale of models and datasets. To mitigate communication overhead, local updates are often applied before global aggregation, resulting in a nested optimization approach with inner and outer steps. However, heavy-tailed stochastic gradient noise remains a significant challenge, particularly in attention-based models, hindering effective training. In this work, we propose TailOPT, an efficient framework designed to address heavy-tailed noise by leveraging adaptive optimization and novel clipping techniques. We establish convergence guarantees for the TailOPT framework under heavy-tailed noise with local updates and potentially unbounded gradient variance. Among its variants, we propose a memory- and communication-efficient instantiation (named $Bi^2Clip$) that performs coordinate-wise clipping from both above and below at both the inner and outer optimizers. $Bi^2Clip$ brings about benefits of adaptive optimization (e.g., Adam) without the cost of maintaining or transmitting additional gradient statistics. Empirically, TailOPT, including $Bi^2Clip$, demonstrates superior performance on various tasks and models compared with state-of-the-art methods, while being more efficient.

Neural network verifiers based on linear bound propagation scale impressively to massive models but can be surprisingly loose when neuron coupling is crucial. Conversely, semidefinite programming (SDP) verifiers capture inter-neuron coupling naturally, but their cubic complexity restricts them to only small models. In this paper, we propose SDP-CROWN, a novel hybrid verification framework that combines the tightness of SDP relaxations with the scalability of bound-propagation verifiers. At the core of SDP-CROWN is a new linear bound---derived via SDP principles---that explicitly captures $\ell_{2}$-norm-based inter-neuron coupling while adding only one extra parameter per layer. This bound can be integrated seamlessly into any linear bound-propagation pipeline, preserving the inherent scalability of such methods yet significantly improving tightness. In theory, we prove that our inter-neuron bound can be up to a factor of $\sqrt{n}$ tighter than traditional per-neuron bounds. In practice, when incorporated into the state-of-the-art $\alpha$-CROWN verifier, we observe markedly improved verification performance on large models with up to 65 thousand neurons and 2.47 million parameters, achieving tightness that approaches that of costly SDP-based methods.

We design improved approximation algorithms for NP-hard graph problems by incorporating predictions (e.g., learned from past data). Our prediction model builds upon and extends the $\varepsilon$-prediction framework by Cohen-Addad, d'Orsi, Gupta, Lee, and Panigrahi (NeurIPS 2024). We consider an edge-based version of this model, where each edge provides two bits of information, corresponding to predictions about whether each of its endpoints belong to an optimal solution. Even with weak predictions where each bit is only $\varepsilon$-correlated with the true solution, this information allows us to break approximation barriers in the standard setting. We develop algorithms with improved approximation ratios for MaxCut, Vertex Cover, Set Cover, and Maximum Independent Set problems (among others). Across these problems, our algorithms share a unifying theme, where we separately satisfy constraints related to high degree vertices (using predictions) and low-degree vertices (without using predictions) and carefully combine the answers.

We study the design of *iterative combinatorial auctions (ICAs)*. The main challenge in this domain is that the bundle space grows exponentially in the number of items. To address this, recent work has proposed machine learning (ML)-based preference elicitation algorithms that aim to elicit only the most critical information from bidders to maximize efficiency. However, while the SOTA ML-based algorithms elicit bidders' preferences via *value queries*, ICAs that are used in practice elicit information via *demand queries*. In this paper, we introduce a novel ML algorithm that provably makes use of the full information from both value and demand queries, and we show via experiments that combining both query types results in significantly better learning performance in practice. Building on these insights, we present MLHCA, a new ML-powered auction that uses value and demand queries. MLHCA significantly outperforms the previous SOTA, reducing efficiency loss by up to a factor 10, with up to 58% fewer queries. Thus, MLHCA achieves large efficiency improvements while also reducing bidders' cognitive load, establishing a new benchmark for both practicability and efficiency. Our code is available at https://github.com/marketdesignresearch/MLHCA.

Subspace optimization algorithms, such as GaLore (Zhao et al., 2024), have gained attention for pre-training and fine-tuning large language models (LLMs) due to their memory efficiency. However, their convergence guarantees remain unclear, particularly in stochastic settings. In this paper, we reveal that GaLore does not always converge to the optimal solution and provide an explicit counterexample to support this finding. We further explore the conditions under which GaLore achieves convergence, showing that it does so when either (i) a sufficiently large mini-batch size is used or (ii) the gradient noise is isotropic. More significantly, we introduce **GoLore** (**G**radient rand**o**m **Lo**w-**r**ank proj**e**ction), a novel variant of GaLore that provably converges in typical stochastic settings, even with standard batch sizes. Our convergence analysis extends naturally to other subspace optimization algorithms. Finally, we empirically validate our theoretical results and thoroughly test the proposed mechanisms. Codes are available at https://github.com/pkumelon/Golore.

Ensuring safety is a critical challenge in applying Reinforcement Learning (RL) to real-world scenarios. Constrained Reinforcement Learning (CRL) addresses this by maximizing returns under predefined constraints, typically formulated as the expected cumulative cost. However, expectation-based constraints overlook rare but high-impact extreme value events in the tail distribution, such as black swan incidents, which can lead to severe constraint violations. To address this issue, we propose the Extreme Value policy Optimization (EVO) algorithm, leveraging Extreme Value Theory (EVT) to model and exploit extreme reward and cost samples, reducing constraint violations. EVO introduces an extreme quantile optimization objective to explicitly capture extreme samples in the cost tail distribution. Additionally, we propose an extreme prioritization mechanism during replay, amplifying the learning signal from rare but high-impact extreme samples. Theoretically, we establish upper bounds on expected constraint violations during policy updates, guaranteeing strict constraint satisfaction at a zero-violation quantile level. Further, we demonstrate that EVO achieves a lower probability of constraint violations than expectation-based methods and exhibits lower variance than quantile regression methods. Extensive experiments show that EVO significantly reduces constraint violations during training while maintaining competitive policy performance compared to baselines.

We explore ensembles of axis-aligned decision stumps, which can be viewed as a generalized additive model (GAM). In this model, stumps utilizing the same feature are grouped to form a shape function for that feature. Instead of relying on boosting or bagging, we employ alternating optimization to learn a fixed-size stump forest. We optimize the parameters of each stump exactly through enumeration, given the other stumps are fixed. For fixed stump splits, the leaf values are optimized jointly by solving a convex problem. To address the overfitting issue inherent in naive optimization of stump forests, we propose effective regularization techniques. Our regularized stump forests achieve accuracy comparable to state-of-the-art GAM methods while using fewer parameters. This work is the first to successfully learn stump forests without employing traditional ensembling techniques like bagging or boosting.

Fine-tuning large-scale pre-trained models is prohibitively expensive in terms of computation and memory costs. Low-Rank Adaptation (LoRA), a popular Parameter-Efficient Fine-Tuning (PEFT) method, offers an efficient solution by optimizing only low-rank matrices. Despite recent progress in improving LoRA's performance, the relationship between the LoRA optimization space and the full parameter space is often overlooked. A solution that appears flat in the loss landscape of the LoRA space may still exhibit sharp directions in the full parameter space, potentially compromising generalization. We introduce Flat-LoRA, which aims to identify a low-rank adaptation situated in a flat region of the full parameter space. Instead of adopting the well-established sharpness-aware minimization approach, which incurs significant computation and memory overheads, we employ a Bayesian expectation loss objective to preserve training efficiency. Further, we design a refined strategy for generating random perturbations to enhance performance and carefully manage memory overhead using random seeds. Experiments across diverse tasks—including mathematical reasoning, coding abilities, dialogue generation, instruction following, and text-to-image generation—demonstrate that Flat-LoRA improves both in-domain and out-of-domain generalization. Code is available at https://github.com/nblt/Flat-LoRA.

We revisit the recently introduced Local Glivenko-Cantelli setting, which studies distribution-dependent uniform convergence rates of the Empirical Mean Estimator (EME). In this work, we investigate generalizations of this setting where arbitrary estimators are allowed rather than just the EME. Can a strictly larger class of measures be learned? Can better risk decay rates be obtained? We provide exhaustive answers to these questions—which are both negative, provided the learner is barred from exploiting some infinite-dimensional pathologies. On the other hand, allowing such exploits does lead to a strictly larger class of learnable measures.