ICLR 2026 论文列表

Scaling multi-task low-rank adaptation (LoRA) to a large number of tasks induces catastrophic performance degradation, such as an accuracy drop from 88.2% to 2.0% on DOTA when scaling from 5 to 15 tasks. This failure is due to parameter and representation misalignment. We find that existing solutions, like regularization and dynamic routing, fail at scale because they are constrained by a fundamental trade-off: strengthening regularization to reduce inter-task conflict inadvertently suppresses the essential feature discrimination required for effective routing. In this work, we identify two root causes for this trade-off. First, uniform regularization disrupts inter-task knowledge sharing: shared underlying knowledge concentrates in high-SV components (89% alignment on Flanv2→BBH). Uniform regularization forces high-SV components to update in orthogonal directions, directly disrupting the shared knowledge. Second, Conflict Amplification: Applying LoRA at the component-level (*e.g.*, $W_q, W_v$) amplifies gradient conflicts; we show block-level adaptation reduces this conflict with 47% fewer parameters. Based on these insights, we propose mtLoRA, a scalable solution with three novel designs: 1) Spectral-Aware Regularization to selectively orthogonalize low-SV components while preserving high-SV shared knowledge, 2) Block-Level Adaptation to mitigate conflict amplification and largely improve parameter efficiency, and 3) Fine-Grained Routing using dimension-specific weights for superior expressive power. On four large-scale (15-25 tasks) vision (DOTA and iNat2018) and NLP (Dolly-15k and BBH) benchmarks, mtLoRA achieves 91.7%, 81.5%, 44.5% and 38.5% accuracy on DOTA, iNat2018, Dolly-15k and BBH respectively, outperforming the state-of-the-art by 2.3% on average while using 47% fewer parameters and 24% less training time.

Clustering is a fundamental primitive in unsupervised learning. However, classical algorithms for $k$-clustering (such as $k$-median and $k$-means) assume access to exact pairwise distances---an unrealistic requirement in many modern applications. We study clustering in the \emph{Rank-model (R-model)}, where access to distances is entirely replaced by a \emph{quadruplet oracle} that provides only relative distance comparisons. In practice, such an oracle can represent learned models or human feedback, and is expected to be noisy and entail an access cost. Given a metric space with $n$ input items, we design randomized algorithms that, using only a noisy quadruplet oracle, compute a set of $O(k \cdot \mathsf{polylog}(n))$ centers along with a mapping from the input items to the centers such that the clustering cost of the mapping is at most constant times the optimum $k$-clustering cost. Our method achieves a query complexity of $O(n\cdot k \cdot \mathsf{polylog}(n))$ for arbitrary metric spaces and improves to $O((n+k^2) \cdot \mathsf{polylog}(n))$ when the underlying metric has bounded doubling dimension. When the metric has bounded doubling dimension we can further improve the approximation from constant to $1+\varepsilon$, for any arbitrarily small constant $\varepsilon\in(0,1)$, while preserving the same asymptotic query complexity. Our framework demonstrates how noisy, low-cost oracles, such as those derived from large language models, can be systematically integrated into scalable clustering algorithms.

Designing efficient optimizers for large language models (LLMs) with low-memory requirements and fast convergence is an important and challenging problem. This paper makes a step towards the systematic design of such optimizers through the lens of structured Fisher information matrix (FIM) approximation. We show that many state-of-the-art efficient optimizers can be viewed as solutions to FIM approximation (under the Frobenius norm) with specific structural assumptions. Building on these insights, we propose two design recommendations of practical efficient optimizers for LLMs, involving the careful selection of structural assumptions to balance generality and efficiency, and enhancing memory efficiency of optimizers with general structures through a novel low-rank extension framework. We demonstrate how to use each design approach by deriving new memory-efficient optimizers: Row and Column Scaled SGD (RACS) and Adaptive low-dimensional subspace estimation (Alice). Experiments on LLaMA pre-training (up to 1B parameters) validate the effectiveness, showing faster and better convergence than existing memory-efficient baselines and Adam with little memory overhead. Notably, Alice achieves better than 2x faster convergence over Adam, while RACS delivers strong performance on the 1B model with SGD-like memory.

The rapid growth in the parameter scale of large language models (LLMs) has created a high demand for efficient compression techniques. As a hardware-agnostic and highly compatible technique, low-rank compression has been widely adopted. However, existing methods typically compress each layer independently by minimizing per-layer reconstruction error, overlooking a critical limitation: the reconstruction error propagates and accumulates through the network, which leads to amplified global deviations from the full-precision baseline. To address this, we propose **Self-Adaptive Error Suppression SVD (SAES-SVD)**, a LLMs compression framework that jointly optimizes intra-layer reconstruction and inter-layer error compensation. SAES-SVD is composed of two novel components: **Cumulative Error-Aware Layer Compression (CEALC),** which formulates the compression objective as a combination of local reconstruction and weighted cumulative error compensation. Based on it, we derive a closed-form low-rank solution relied on second-order activation statistics, which explicitly aligns each layer's output with its full-precision counterpart to compensate for accumulated errors. \ding{183} Adaptive Collaborative Error Suppression (ACES), which automatically adjusts the weighting coefficient to enhance the low-rank structure of the compression objective in CELAC. Specifically, the coefficient is optimized to maximize the ratio between the Frobenius norm of the compressed layer's output and that of the compression objective under a fixed rank, thus ensuring that the rank budget is utilized effectively. Extensive experiments across multiple LLM architectures and tasks show that, without fine-tuning or additional tricks, SAES-SVD consistently improves post-compression performance. For example, at a 0.2 compression ratio on LLaMA-7B, existing methods exhibit an average accuracy drop exceeding 0.05, whereas SAES-SVD restricts the drop to only 0.02. These improvements underscore the potential of SAES-SVD to effectively narrow the gap between compressed models and their full-precision counterparts, paving the way for more reliable compression of LLMs.

Fully polynomial neural networks—models whose computations comprise only additions and multiplications—are attractive for privacy-preserving inference under homomorphic encryption (HE). Yet most prior systems obtain such models by *post-hoc* replacement of nonlinearities with high-degree or cascaded polynomials, which inflates HE cost and makes training numerically fragile and hard to scale. We introduce **ULD-Net**, a pretraining methodology that enables *ultra-low-degree* (multiplicative depth ≤ 3 for each operator) fully polynomial networks to be trained from scratch at ImageNet and transformer scale while maintaining high accuracy. The key is a polynomial-only normalization, **PolyNorm**, coupled with a principled choice of normalization axis that keeps activations in a well-conditioned range across deep stacks of polynomial layers. Together with a special set of polynomial-aware operator replacements, such as polynomial activation functions and linear attention, ULD-Net delivers stable optimization without resorting to high-degree approximations. Experimental results demonstrate that ULD-Net outperforms several state-of-the-art open-source fully and partially polynomial approaches across both CNNs and ViTs on diverse datasets, in terms of both accuracy and HE inference latency. Specifically, ULD-Net achieves +0.39% accuracy and a 2.76× speedup compared to the best fully polynomial baseline; up to +3.33% accuracy and a 3.17× speedup compared to the best partial polynomial baseline. Applying ULD-Net to ViT-Small and ViT-Base yields 76.7% and 75.2% top-1 accuracy on ImageNet, demonstrating the first fully polynomial models scaled to the ViT/ImageNet level.

The *Learnware* paradigm aims to help users solve machine learning tasks by leveraging existing well-trained models rather than starting from scratch. A learnware comprises a submitted model paired with a *specification* sketching its capabilities. For an open platform with continuously uploaded models, these specifications are essential to enabling users to identify helpful models, eliminating the requirement for prohibitively costly per-model evaluations. In previous research, specifications based on privacy-preserving reduced sets succeed in enabling learnware identification through distribution matching, but suffer from high sample complexity for learnwares from high-dimensional, unstructured data like images or text. In this paper, we formalize **Pa**rameter **Ve**ctor (PAVE) specification for learnware identification, which utilizes the changes in pre-trained model parameters to inherently encode the model capability and task requirements, offering an effective solution for these learnwares. Theoretically, from the neural tangent kernel perspective, we establish a tight connection between PAVE and prior specifications, providing a theoretical explanation for their shared underlying principles. We further approximate the parameter vector in a low-rank space and analyze the approximation error bound, highly reducing the computational and storage overhead. Extensive empirical studies demonstrate that PAVE specification excels at identifying CV and NLP learnwares for reuse on given user tasks, and succeeds in identifying helpful learnwares from open learnware repository with corrupted model quality for the first time. Reusing identified learnware to solve user tasks can even outperform user-fine-tuned pre-trained models in data-limited scenarios.

Time series forecasting remains a critical challenge across numerous domains, yet the effectiveness of complex models often varies unpredictably across datasets. Recent studies highlight the surprising competitiveness of simple linear models, suggesting that their robustness and interpretability warrant deeper theoretical investigation. This paper presents a systematic study of linear models for time series forecasting, with a focus on the role of characteristic roots in temporal dynamics. We begin by analyzing the noise-free setting, where we show that characteristic roots govern long-term behavior and explain how design choices such as instance normalization and channel independence affect model capabilities. We then extend our analysis to the noisy regime, revealing that models tend to produce spurious roots. This leads to the identification of a key data-scaling property: mitigating the influence of noise requires disproportionately large training data, highlighting the need for structural regularization. To address these challenges, we propose two complementary strategies for robust root restructuring. The first uses rank reduction techniques, including \textbf{Reduced-Rank Regression (RRR)} and \textbf{Direct Weight Rank Reduction (DWRR)}, to recover the low-dimensional latent dynamics. The second, a novel adaptive method called \textbf{Root Purge}, encourages the model to learn a noise-suppressing null space during training. Extensive experiments on standard benchmarks demonstrate the effectiveness of both approaches, validating our theoretical insights and achieving state-of-the-art results in several settings. Our findings underscore the potential of integrating classical theories for linear systems with modern learning techniques to build robust, interpretable, and data-efficient forecasting models. The code is publicly available at: https://github.com/Wangzzzzzzzz/Root-Purge-for-Time-Series-Forecasting.

Regret Matching (RM) and its variants are widely employed to learn a Nash equilibrium (NE) in large-scale games. However, most existing research only establishes a theoretical convergence rate of $O(1/\sqrt{T})$ for these algorithms in learning an NE. Recent studies have shown that smooth RM$^+$ variants, the advanced variants of RM, can achieve an improved convergence rate of $O(1/T)$. Despite this improvement, smooth RM$^+$ variants lose the parameter-free property, i.e., no parameters that need to be tuned, a highly desirable feature in practical applications. In this paper, we propose a novel smooth RM$^+$ variant called Monotone Increasing Smooth Predictive Regret Matching$^+$ (MI-SPRM$^+$), which retains the parameter-free property while still achieving a theoretical convergence rate of $O(1/T)$. To achieve these properties, MI-SPRM$^+$ employs a technology called Adaptive Regret Domain (ARD), which ensures that the lower bound for the 1-norm of accumulated regrets increases monotonically by adjusting the decision space at each iteration. This design is motivated by the observation that the range of step-sizes supporting the $O(1/T)$ convergence rate in existing smooth RM$^+$ variants is contingent on the lower bound for the 1-norm of accumulated regrets. Experimental results confirm that MI-SPRM$^+$ empirically attains an $O(1/T)$ convergence rate.

Mixture-of-Experts (MoE) architectures in large language models (LLMs) deliver exceptional performance and reduced inference costs compared to dense LLMs. However, their large parameter counts result in prohibitive memory requirements, limiting practical deployment. While existing pruning methods primarily focus on expert-level pruning, this coarse granularity often leads to substantial accuracy degradation. In this work, we introduce HEAPr, a novel pruning algorithm that decomposes experts into smaller, indivisible atomic experts, enabling more precise and flexible atomic expert pruning. To measure the importance of each atomic expert, we leverage second-order information based on principles similar to Optimal Brain Surgeon (OBS) theory. To address the computational and storage challenges posed by second-order information, HEAPr exploits the inherent properties of atomic experts to transform the second-order information from expert parameters into that of atomic expert parameters, and further simplifies it to the second-order information of atomic expert outputs. This approach reduces the space complexity from $O(d^4)$, where $d$ is the model’s dimensionality, to $O(d^2)$. HEAPr requires only two forward passes and one backward pass on a small calibration set to compute the importance of atomic experts. Extensive experiments on MoE models, including DeepSeek MoE and Qwen MoE family, demonstrate that HEAPr outperforms existing expert-level pruning methods across a wide range of compression ratios and benchmarks. Specifically, HEAPr achieves nearly lossless compression at compression ratios of $20\% \sim 25\%$ in most models, while also reducing FLOPs nearly by $20\%$. The code can be found at \href{https://anonymous.4open.science/r/anonymous-code-B927/}{anonymous-code-B927}.

A fundamental challenge in the theory of deep learning is to understand whether gradient-based training can promote parameters belonging to certain lower-dimensional structures (e.g., sparse or low-rank sets), leading to so-called implicit bias. As a stepping stone, motivated by the proof structure of existing intrinsic bias analyses, we study when a gradient flow on a parameter $\theta$ implies an intrinsic gradient flow on a ``lifted'' variable $z = \phi(\theta)$, for an architecture-related function $\phi$. We express a so-called intrinsic dynamic property and show how it is related to the study of conservation laws associated with the factorization $\phi$. This leads to a simple criterion based on the inclusion of kernels of linear maps, which yields a necessary condition for this property to hold. We then apply our theory to general ReLU networks of arbitrary depth and show that, for a dense set of initializations, it is possible to rewrite the flow as an intrinsic dynamic in a lower dimension that depends only on $z$ and the initialization, when $\phi$ is the so-called path-lifting. In the case of linear networks with $\phi$, the product of weight matrices, the intrinsic dynamic is known to hold under so-called balanced initializations; we generalize this to a broader class of {\em relaxed balanced} initializations, showing that, in certain configurations, these are the \emph{only} initializations that ensure the intrinsic metric property. Finally, for the linear neural ODE associated with the limit of infinitely deep linear networks, with relaxed balanced initialization, we explicit the corresponding intrinsic dynamics.

As large pre-trained models rapidly scale, Parameter-Efficient Fine-Tuning (PEFT) through methods like Low-Rank Adaptation (LoRA) becomes increasingly crucial. While LoRA has emerged as a cornerstone of PEFT, excelling at preserving performance with minimal additional parameters, exploring parameter-sharing mechanisms of LoRA remains critical to pushing efficiency boundaries. However, existing naive LoRA sharing methods often degrade performance due to sacrificed representational diversity and weakened model expressiveness. To overcome this issue, we conduct an in-depth analysis of pre-trained models using gradient-based proxy entropy, and uncover two critical, previously overlooked properties: Local Similarity and Layer-wise Information Heterogeneity. Building on these insights, we propose E²LoRA, a novel dual-adaptive sharing framework. It enables adaptive sharing interval partitioning, guided by inter-layer proxy entropy similarity, and adaptive rank allocation, informed by layer-wise absolute proxy entropy. This unique design leverages inherently informative properties of pre-trained models to significantly reduce parameter redundancy while maintaining or enhancing expressiveness. Comprehensive evaluations across diverse tasks, modalities, and models consistently demonstrate that E²LoRA achieves an excellent balance of efficiency and effectiveness, consistently matching or surpassing baselines with approximately 50% fewer trainable parameters.

Numerous studies on low-rank adaptation (LoRA) emerged in recent years, with the aim of accelerating the convergence of the LoRA framework. In this paper, we leverage the horizontal lift theory from differential geometry to establish the general iteration scheme on the quotient manifold \mathbb{R}\_\*^{m \times r} \times \mathbb{R}\_\*^{n \times r}/\sim. By endowing the LoRA framework with Riemannian quotient geometries, our theory not only guarantees efficient feature learning but also bridges the LoRA algorithms and the pre-training algorithms for large models. Furthermore, we theoretically analyze the role of the weight decay matrix $\epsilon_{decay}I$ in efficient feature learning and then replace it with the Sylvester matrix $K$, indicating that the theory helps remove an important hyperparameter while generating accurate and computationally efficient optimizers. Based on the general scheme, we propose two efficient LoRA optimizers with runtime analysis, Adam-Sylvester (AdamS) and LRACS, then conduct experiments on the transformer-based networks. The results demonstrate evident improvements over existing optimizers.

Recent attempts to combine low-rank adaptation (LoRA) with mixture-of-experts (MoE) for multi-task adaptation of Large Language Models (LLMs) often replace whole attention/FFN layers with switch experts or append parallel expert branches, undermining parameter efficiency and limiting task specialization. We introduce **LoRA-Mixer**, a modular MoE framework that routes task-specific LoRA experts into the core projection matrices of the attention module (input/output linear layers), rather than primarily targeting FFN blocks. The design delivers fine-grained token-level specialization by fully exploiting the attention mechanism, while remaining drop-in compatible with Transformers and state-space models (SSMs) as the linear projection layers are ubiquitous. To train robust routers from limited data while promoting stable, selective decisions and high expert reuse, **LoRA-Mixer** employs an adaptive **Routing Specialization Loss (RSL)** that jointly enforces global load balance and input-aware specialization via an entropy-shaping objective. The framework supports two regimes: (i) joint optimization of adapters and router with a differentiable hard–soft top-k routing scheme, and (ii) plug-and-play routing over frozen, pre-trained LoRA modules sourced from public repositories. Across 15 benchmarks—including MedQA, GSM8K, HumanEval, and GLUE—RSL-optimized LoRA-Mixer outperforms state-of-the-art routing and LoRA-MoE baselines while using 48% of their trainable parameters, with gains of +3.79%, +2.90%, and +3.95% on GSM8K, CoLA, and ARC-C, respectively. Cross-model transfer and adapter reuse experiments further demonstrate the approach’s versatility and data efficiency.

Amortized simulator-based inference offers a powerful framework for tackling Bayesian inference in computational fields such as engineering or neuroscience, increasingly leveraging modern generative methods like diffusion models to map observed data to model parameters or future predictions. These approaches yield posterior or posterior-predictive samples for new datasets without requiring further simulator calls after training on simulated parameter-data pairs. However, their applicability is often limited by the prior distribution(s) used to generate model parameters during this training phase. To overcome this constraint, we introduce *PriorGuide*, a technique specifically designed for diffusion-based amortized inference methods. PriorGuide leverages a novel guidance approximation that enables flexible adaptation of the trained diffusion model to new priors at test time, crucially without costly retraining. This allows users to readily incorporate updated information or expert knowledge post-training, enhancing the versatility of pre-trained inference models.

Generative models have demonstrated significant results in ultra-low bitrate image compression, owing to their powerful capabilities for content generation and texture completion. Existing works primarily based on diffusion models still face challenges such as limited bitrate adaptability and high computational complexity for encoding and decoding. Inspired by the success of Visual AutoRegressive model (VAR), we introduce AutoRegressive-based Progressive Coding (ARPC) for ultra-low bitrate image compression, a progressive image compression framework based on next-scale prediction visual autoregressive model. Based on multi-scale residual vector quantizer, ARPC efficiently encodes the image into multi-scale discrete token maps and controls the bitrates by selecting different scales for transmission. For decompression, ARPC leverages the prior knowledge inherent in the visual autoregressive model to predict the unreceived scales, which is naturally the autoregressive generation process. To further increase the compression ratio, we target the VAR as a probability estimator for lossless entropy coding and propose group-masked bitwise multi-scale residual quantizer to adaptively allocate bits for different scales. Extensive experiments show that ARPC achieves state-of-the-art perceptual fidelity at ultra-low bitrates and high decompression efficiency compared with existing diffusion-based methods.

Driven by the rapid growth of model parameters, parameter-efficient fine-tuning (PEFT) has become essential for adapting large models to diverse downstream tasks under constrained computational resources. Within this paradigm, orthogonal fine-tuning and its variants preserve semantic representations of pre-trained models, but struggle to achieve both expressiveness and efficiency in terms of parameter counts, memory, and computation. To overcome this limitation, we propose efficient Orthogonal Fine-Tuning with Principal Subspace adaptation (PSOFT), which confines orthogonal transformations to the principal subspace of pre-trained weights. Specifically, PSOFT constructs this subspace via matrix decomposition to enable compatible transformations with higher rank, establishes a theoretical condition that strictly maintains the geometry of this subspace for essential semantic preservation, and introduces efficient tunable vectors that gradually relax orthogonality during training to enhance adaptability. Extensive experiments on 35 NLP and CV tasks across four representative models demonstrate that PSOFT offers a practical and scalable solution to simultaneously achieve semantic preservation, expressiveness, and multi-dimensional efficiency in PEFT.

Reinforcement learning (RL) has become a cornerstone for enhancing the reasoning capabilities of large language models (LLMs), with recent innovations such as Group Relative Policy Optimization (GRPO) demonstrating exceptional effectiveness. In this study, we identify a critical yet underexplored issue in RL training: low-probability tokens disproportionately influence model updates due to their large gradient magnitudes. This dominance hinders the effective learning of high-probability tokens, whose gradients are essential for LLMs' performance but are substantially suppressed. To mitigate this interference, we propose two novel methods: Advantage Reweighting and Low-Probability Token Isolation (Lopti), both of which effectively attenuate gradients from low-probability tokens while emphasizing parameter updates driven by high-probability tokens. Our approaches promote balanced updates across tokens with varying probabilities, thereby enhancing the efficiency of RL training. Experimental results demonstrate that they substantially improve the performance of GRPO-trained LLMs, achieving up to a 46.2% improvement in K&K Logic Puzzle reasoning tasks.

Fine-tuning safety-aligned large language models (LLMs) can substantially compromise their safety. Previous approaches require many safety samples or calibration sets, which not only incur significant computational overhead during realignment but also lead to noticeable degradation in model utility. Contrary to this belief, we show that safety alignment can be fully recovered with only a single safety example, without sacrificing utility and at minimal cost. Remarkably, this recovery is effective regardless of the number of harmful examples used in fine-tuning or the size of the underlying model, and convergence is achieved within just a few epochs. Furthermore, we uncover the low-rank structure of the safety gradient, which explains why such efficient correction is possible. We validate our findings across five safety-aligned LLMs and multiple datasets, demonstrating the generality of our approach.

Visual Prompt Tuning (VPT) has proven effective for parameter-efficient adaptation of pre-trained vision models to downstream tasks by inserting task-specific learnable prompt tokens. Despite its empirical success, a comprehensive theoretical understanding of VPT remains an active area of research. Building on the recently established connection between Mixture of Experts (MoE) and prompt-based methods, wherein each attention head can be conceptualized as a composition of multiple MoE models, we reinterpret VPT as the introduction of new *prompt experts* into these MoE structures. We identify a key limitation in existing VPT frameworks: the *restricted functional expressiveness* of prompt experts, which remain static and thus limited in their adaptability. To address this, we propose **Visual Adaptive Prompt Tuning (VAPT)**, a novel method that endows prompt experts with enhanced expressiveness while preserving parameter efficiency. Empirical evaluations on VTAB-1K and FGVC demonstrate that VAPT achieves *substantial performance improvements*, surpassing fully fine-tuned baselines by **7.34%** and **1.04%**, respectively. Moreover, VAPT consistently outperforms VPT while *requiring fewer additional parameters*. Furthermore, our theoretical analysis indicates that VAPT achieves optimal sample efficiency. Collectively, these results underscore the theoretical grounding and empirical advantages of our approach.