低秩适配

Continual multimodal instruction tuning is crucial for adapting Multimodal Large Language Models (MLLMs) to evolving tasks. However, most existing methods adopt a fixed architecture, struggling with adapting to new tasks due to static model capacity. We propose to evolve the architecture under parameter budgets for dynamic task adaptation, which remains unexplored and imposes two challenges: 1) task architecture conflict, where different tasks require varying layer-wise adaptations, and 2) modality imbalance, where different tasks rely unevenly on modalities, leading to unbalanced updates. To address these challenges, we propose a novel Dynamic Mixture of Curriculum LoRA Experts (D-MoLE) method, which automatically evolves MLLM's architecture with controlled parameter budgets to continually adapt to new tasks while retaining previously learned knowledge. Specifically, we propose a dynamic layer-wise expert allocator, which automatically allocates LoRA experts across layers to resolve architecture conflicts, and routes instructions layer-wisely to facilitate knowledge sharing among experts. Then, we propose a gradient-based inter-modal continual curriculum, which adjusts the update ratio of each module in MLLM based on the difficulty of each modality within the task to alleviate the modality imbalance problem. Extensive experiments show that D-MoLE significantly outperforms state-of-the-art baselines, achieving a 15 percent average improvement over the best baseline. To the best of our knowledge, this is the first study of continual learning for MLLMs from an architectural perspective.

Prompt learning is a cutting-edge parameter-efficient fine-tuning technique for pre-trained vision-language models (VLMs). Instead of learning a single text prompt, recent works have revealed that learning diverse text prompts can effectively boost the performances on downstream tasks, as the diverse prompted text features can comprehensively depict the visual concepts from different perspectives. However, diverse prompt learning demands enormous computational resources. This efficiency issue still remains unexplored. To achieve efficient and diverse prompt learning, this paper proposes a novel \textbf{Surrogate Prompt Learning (SurPL)} framework. Instead of learning diverse text prompts, SurPL directly generates the desired prompted text features via a lightweight \textbf{Surrogate Feature Generator (SFG)}, thereby avoiding the complex gradient computation procedure of conventional diverse prompt learning. Concretely, based on a basic prompted text feature, SFG can directly and efficiently generate diverse prompted features according to different pre-defined conditional signals. Extensive experiments indicate the effectiveness of the surrogate prompted text features, and show compelling performances and efficiency of SurPL on various benchmarks.

Adapting pre-trained large language models (LLMs) is crucial but challenging due to their enormous size. Parameter-efficient fine-tuning (PEFT) techniques typically employ additive adapters applied to frozen model weights. To further reduce memory usage, model weights are often compressed through quantization. However, existing PEFT methods often yield suboptimal model quality because they rely on restrictive assumptions, such as low-rank constraints on adapters to limit the number of trainable parameters. We find that sketching, a popular data compression technique, can serve as an efficient LLM adaptation strategy while avoiding the low-rank assumption. We introduce SketchTune, a compressive adaptation strategy that compresses LLM weights into compact fine-tunable sketches, integrating compression and adaptation into a unified framework. This integration eliminates the need for complex two-path computation in existing PEFT techniques, enabling faster and more memory-efficient training and inference. SketchTune is supported by mathematical insights into matrix classes that are better approximated using sketching rather than low-rank methods. Our extensive evaluations with Llama and Mistral models demonstrate that SketchTune outperforms leading PEFT methods across diverse tasks while using substantially smaller base models and comparable trainable parameters. As a highlight, SketchTune outperforms LoRA, DoRA, and S2FT on commonsense and math benchmarks using 2.6-3.5$\times$ smaller base models and exceeds LoftQ in accuracy by 14.48\% on GSM8K with 7.3$\times$ fewer trainable parameters.

While Foundation Models provide a general tool for rapid content creation, they regularly require task-specific adaptation. Traditionally, this exercise involves careful curation of datasets and repeated fine-tuning of the underlying model. Fine-tuning techniques enable practitioners to adapt foundation models for many new applications but require expensive and lengthy training while being notably sensitive to hyperparameter choices. To overcome these limitations, we introduce Text-to-LoRA (T2L), a model capable of adapting large language models (LLMs) on the fly solely based on a natural language description of the target task. T2L is a hypernetwork trained to construct LoRAs in a single inexpensive forward pass. After training T2L on a suite of 9 pre-trained LoRA adapters (GSM8K, Arc, etc.), we show that the ad-hoc reconstructed LoRA instances match the performance of task-specific adapters across the corresponding test sets. Furthermore, T2L can compress hundreds of LoRA instances and zero-shot generalize to entirely unseen tasks. This approach provides a significant step towards democratizing the specialization of foundation models and enables language-based adaptation with minimal compute requirements. Our code is available at https://github.com/SakanaAI/text-to-lora

Fine-tuning LLMs is both computationally and memory-intensive. While parameter-efficient fine-tuning methods, such as QLoRA and DoRA, reduce the number of trainable parameters and lower memory usage, they do not decrease computational cost. In some cases, they may even slow down fine-tuning. In this paper, we introduce SparseLoRA, a method that accelerates LLM fine-tuning through contextual sparsity. We propose a lightweight, training-free SVD sparsity estimator that dynamically selects a sparse subset of weights for loss and gradient computation. Also, we systematically analyze and address sensitivity across layers, tokens, and training steps. Our experimental results show that SparseLoRA reduces computational cost by up to $2.0\times$ and a measured speedup of up to $1.5\times$ while maintaining accuracy across various downstream tasks, including commonsense and arithmetic reasoning, code generation, and instruction following.

In order to streamline the fine-tuning of foundation models, Low-Rank Adapters (LoRAs) have been substantially adopted across various fields, including instruction tuning and domain adaptation. The underlying concept of LoRA involves decomposing a full-rank matrix into the product of two lower-rank matrices, which reduces storage consumption and accelerates the training process. Furthermore, to address the limited expressive capacity of LoRA, the Mixture-of-Expert (MoE) has been introduced for incorporating multiple LoRA adapters. The integration of LoRA experts leads to a visible improvement across several downstream scenes. However, the mixture of LoRAs (MoE-LoRA) still exhibits its low robustness during tuning and inferring. Inspired by the Riemannian Preconditioners which train LoRA as a sub-space projector, we propose a new training strategy for MoE-LoRA, to stabilize and boost its feature learning by gate-rescaled multi-space projections. We provide both a theoretical solution as well as an alternative engineering strategy. Examinations on SGD and AdamW optimizers demonstrate the effectiveness of our methodology. Source code is available at https://github.com/THUDM/MoELoRA_Riemannian.

We introduce a novel generative framework for functions by integrating Implicit Neural Representations (INRs) and Transformer-based hypernetworks into latent variable models. Unlike prior approaches that rely on MLP-based hypernetworks with scalability limitations, our method employs a Transformer-based decoder to generate INR parameters from latent variables, addressing both representation capacity and computational efficiency. Our framework extends latent diffusion models (LDMs) to INR generation by replacing standard decoders with a Transformer-based hypernetwork, which can be trained either from scratch or via hyper-transforming—a strategy that fine-tunes only the decoder while freezing the pre-trained latent space. This enables efficient adaptation of existing generative models to INR-based representations without requiring full retraining. We validate our approach across multiple modalities, demonstrating improved scalability, expressiveness, and generalization over existing INR-based generative models. Our findings establish a unified and flexible framework for learning structured function representations.

The LoRA method has achieved notable success in reducing GPU memory usage by applying low-rank updates to weight matrices. Yet, one simple question remains: can we push this reduction even further? Furthermore, is it possible to achieve this while improving performance and reducing computation time? Answering these questions requires moving beyond the conventional weight-centric approach. In this paper, we present a state-based fine-tuning framework that shifts the focus from weight adaptation to optimizing forward states, with LoRA acting as a special example. Specifically, state-based tuning introduces parameterized perturbations to the states within the computational graph, allowing us to control states across an entire residual block. A key advantage of this approach is the potential to avoid storing large intermediate states in models like transformers. Empirical results across multiple architectures—including ViT, RoBERTa, LLaMA2-7B, and LLaMA3-8B—show that our method further reduces memory consumption and computation time while simultaneously improving performance. Moreover, as a result of memory reduction, we explore the feasibility to train 7B/8B models on consumer-level GPUs like Nvidia 3090, without model quantization. The code is available at an anonymous GitHub repository

Is there really much more to say about sparse autoencoders (SAEs)? Autoencoders in general, and SAEs in particular, represent deep architectures that are capable of modeling low-dimensional latent structure in data. Such structure could reflect, among other things, correlation patterns in large language model activations, or complex natural image manifolds. And yet despite the wide-ranging applicability, there have been relatively few changes to SAEs beyond the original recipe from decades ago, namely, standard deep encoder/decoder layers trained with a classical/deterministic sparse regularizer applied within the latent space. One possible exception is the variational autoencoder (VAE), which adopts a stochastic encoder module capable of producing sparse representations when applied to manifold data. In this work we formalize underappreciated weaknesses with both canonical SAEs, as well as analogous VAEs applied to similar tasks, and propose a hybrid alternative model that circumvents these prior limitations. In terms of theoretical support, we prove that global minima of our proposed model recover certain forms of structured data spread across a union of manifolds. Meanwhile, empirical evaluations on synthetic and real-world datasets substantiate the efficacy of our approach in accurately estimating underlying manifold dimensions and producing sparser latent representations without compromising reconstruction error. In general, we are able to exceed the performance of equivalent-capacity SAEs and VAEs, as well as recent diffusion models where applicable, within domains such as images and language model activation patterns.

Mixture-of-Experts (MoE) activates only a subset of experts during inference, allowing the model to maintain low inference FLOPs and latency even as the parameter count scales up. However, since MoE dynamically selects the experts, all the experts need to be loaded into VRAM. Their large parameter size still limits deployment, and offloading, which load experts into VRAM only when needed, significantly increase inference latency. To address this, we propose Mixture of Lookup Experts (MoLE), a new MoE architecture that is efficient in both communication and VRAM usage. In MoLE, the experts are Feed-Forward Networks (FFNs) during training, taking the output of the embedding layer as input. Before inference, these experts can be re-parameterized as lookup tables (LUTs) that retrieves expert outputs based on input ids, and offloaded to storage devices. Therefore, we do not need to perform expert computations during inference. Instead, we directly retrieve the expert's computation results based on input ids and load them into VRAM, and thus the resulting communication overhead is negligible. Experiments show that, with the same FLOPs and VRAM usage, MoLE achieves inference speeds comparable to dense models and significantly faster than MoE with experts offloading, while maintaining performance on par with MoE. Code: https://github.com/JieShibo/MoLE.

Modern methods for fine-tuning a Vision Transformer (ViT) like Low-Rank Adaptation (LoRA) and its variants demonstrate impressive performance. However, these methods ignore the high-dimensional nature of Multi-Head Attention (MHA) weight tensors. To address this limitation, we propose Canonical Rank Adaptation (CaRA). CaRA leverages tensor mathematics, first by tensorising the transformer into two different tensors; one for projection layers in MHA and the other for feed-forward layers. Second, the tensorised formulation is fine-tuned using the low-rank adaptation in Canonical-Polyadic Decomposition (CPD) form. Employing CaRA efficiently minimizes the number of trainable parameters. Experimentally, CaRA outperforms existing Parameter-Efficient Fine-Tuning (PEFT) methods in visual classification benchmarks such as Visual Task Adaptation Benchmark (VTAB)-1k and Fine-Grained Visual Categorization (FGVC).

Open-Set Domain Adaptation (OSDA) aims to transfer knowledge from the labeled source domain to the unlabeled target domain that contains unknown categories, thus facing the challenges of domain shift and unknown category recognition. While recent works have demonstrated the potential of causality for domain alignment, little exploration has been conducted on causal-inspired theoretical frameworks for OSDA. To fill this gap, we introduce the concept of *Susceptibility* and propose a novel **C**ounterfactual-based susceptibility risk framework for **OSDA**, termed **COSDA**. Specifically, COSDA consists of three novel components: (i) a *Susceptibility Risk Estimator (SRE)* for capturing causal information, along with comprehensive derivations of the computable theoretical upper bound, forming a risk minimization framework under the OSDA paradigm; (ii) a *Contrastive Feature Alignment (CFA)* module, which is theoretically proven based on mutual information to satisfy the *Exogeneity* assumption and facilitate cross-domain feature alignment; (iii) a *Virtual Multi-unknown-categories Prototype (VMP)* pseudo-labeling strategy, providing label information by measuring how similar samples are to known and multiple virtual unknown category prototypes, thereby assisting in open-set recognition and intra-class discriminative feature learning. Extensive experiments demonstrate that our approach achieves state-of-the-art performance.

The emergence of multi-dimensional data presents significant challenges for traditional regression models based on matrices or vectors, particularly in capturing multi-directional correlations. In response, tensor regression has been proposed as a powerful framework for modeling linear relationships among multi-dimensional variables. In this paper, we introduce a high-dimensional tensor-response tensor regression model under low-dimensional structural assumptions, such as sparsity and low-rankness. Assuming the underlying tensor lies within an unknown low-dimensional subspace, we consider a least squares estimation framework with non-convex penalties. Theoretically, we derive general risk bounds for the resulting estimators and demonstrate that they achieve the oracle statistical rates under mild technical conditions. To compute the proposed estimators efficiently, we introduce an accelerated proximal gradient algorithm demonstrating rapid convergence in practice. Extensive experiments on synthetic and real-world datasets validate the effectiveness of the proposed regression model and showcase the practical utility of the theoretical findings.

The growing computational demands of training large language models (LLMs) necessitate more efficient methods. Quantized training presents a promising solution by enabling low-bit arithmetic operations to reduce these costs. While FP8 precision has demonstrated feasibility, leveraging FP4 remains a challenge due to significant quantization errors and limited representational capacity. This work introduces the first FP4 training framework for LLMs, addressing these challenges with two key innovations: a differentiable quantization estimator for precise weight updates and an outlier clamping and compensation strategy to prevent activation collapse. To ensure stability, the framework integrates a mixed-precision training scheme and vector-wise quantization. Experimental results demonstrate that our FP4 framework achieves accuracy comparable to BF16 and FP8, with minimal degradation, scaling effectively to 13B-parameter LLMs trained on up to 100B tokens. With the emergence of next-generation hardware supporting FP4, our framework sets a foundation for efficient ultra-low precision training.

Machine learning has demonstrated transformative potential for database operations, such as query optimization and in-database data analytics. However, dynamic database environments, characterized by frequent updates and evolving data distributions, introduce concept drift, which leads to performance degradation for learned models and limits their practical applicability. Addressing this challenge requires efficient frameworks capable of adapting to shifting concepts while minimizing the overhead of retraining or fine-tuning. In this paper, we propose FLAIR, an online adaptation framework that introduces a new paradigm called \textit{in-context adaptation} for learned database operations. FLAIR leverages the inherent property of data systems, i.e., immediate availability of execution results for predictions, to enable dynamic context construction. By formalizing adaptation as $f:(\mathbf{x} | \mathcal{C}_t) \to \mathbf{y}$, with $\mathcal{C}_t$ representing a dynamic context memory, FLAIR delivers predictions aligned with the current concept, eliminating the need for runtime parameter optimization. To achieve this, FLAIR integrates two key modules: a Task Featurization Module for encoding task-specific features into standardized representations, and a Dynamic Decision Engine, pre-trained via Bayesian meta-training, to adapt seamlessly using contextual information at runtime. Extensive experiments across key database tasks demonstrate that FLAIR outperforms state-of-the-art baselines, achieving up to $5.2\times$ faster adaptation and reducing error by 22.5\% for cardinality estimation.

In this paper, we present a novel approach, termed Double-Filter,to “slim down” the fine-tuning process of vision-language pre-trained (VLP) models via filtering redundancies in feature inputs and architectural components. We enhance the fine-tuning process using two approaches. First, we develop a new patch selection method incorporating image patch filtering through background and foreground separation, followed by a refined patch selection process. Second, we design a genetic algorithm to eliminate redundant fine-grained architecture layers, improving the efficiency and effectiveness of the model. The former makes patch selection semantics more comprehensive, improving inference efficiency while ensuring semantic representation. The latter’s fine-grained layer filter removes architectural redundancy to the extent possible and mitigates the impact on performance. Experimental results demonstrate that the proposed Double-Filter achieves superior efficiency of model fine-tuning and maintains competitive performance compared with the advanced efficient fine-tuning methods on three downstream tasks, VQA, NLVR and Retrieval. In addition, it has been proven to be effective under METER and ViLT VLP models.

Autoregressive models have emerged as a powerful generative paradigm for visual generation. The current de-facto standard of next token prediction commonly operates over a single-scale sequence of dense image tokens, and is incapable of utilizing global context especially for early tokens prediction. In this paper, we introduce a new autoregressive design to model a hierarchy from a few low-resolution image tokens to the typical dense image tokens, and delve into a thorough hierarchical dependency across multi-scale image tokens. Technically, we present a Hierarchical Masked Autoregressive models (Hi-MAR) that pivot on low-resolution image tokens to trigger hierarchical autoregressive modeling in a multi-phase manner. Hi-MAR learns to predict a few image tokens in low resolution, functioning as intermediary pivots to reflect global structure, in the first phase. Such pivots act as the additional guidance to strengthen the next autoregressive modeling phase by shaping global structural awareness of typical dense image tokens. A new Diffusion Transformer head is further devised to amplify the global context among all tokens for mask token prediction. Extensive evaluations on both class-conditional and text-to-image generation tasks demonstrate that Hi-MAR outperforms typical AR baselines, while requiring fewer computational costs.

Attention layers are the core component of transformers, the current state-of-the-art neural network architecture. Alternatives to softmax-based attention are being explored due to its tendency to hinder effective information flow. Even *at initialisation*, it remains poorly understood why the propagation of signals and gradients through these random networks can be pathological, resulting in issues known as (i) vanishing/exploding gradients and (ii) rank collapse *in depth*, i.e. when all tokens converge to a single representation along layers. While rank collapse in depth naturally arises from repeated matrix multiplications---a common pattern across various architectures---we identify an additional and previously unknown challenge unique to softmax attention layers: (iii) rank collapse *in width*, which occurs as the context length increases. Using Random Matrix Theory, we conduct a rigorous analysis that uncovers a spectral gap between the two largest singular values of the attention matrix as the cause of (iii), which in turn exacerbates (i) and (ii). Building on this insight, we propose a novel yet simple practical solution to mitigate rank collapse in width by removing the outlier eigenvalue(s). Our theoretical framework offers a fresh perspective on recent practical studies, such as (Ye et al., 2024; Ali et al., 2023), whose ad hoc solutions can now be interpreted as implicit efforts to address the spectral gap issue. This work provides valuable theoretical support for ongoing large-scale empirical research, bringing theory and practice one step closer in the understanding of transformers.

Vision GNN (ViG) demonstrates superior performance by representing images as graph structures, providing a more natural way to capture irregular semantic patterns beyond traditional grid or sequence-based representations. To efficiently adapt ViG to downstream tasks, parameter-efficient fine-tuning techniques like visual prompting become increasingly essential. However, existing prompting methods are primarily designed for Transformer-based models, neglecting the rich topological relationships among nodes and edges in graph-based representations, limiting their capacity to model complex semantics. In this paper, we propose Vision Graph Prompting (VGP), a novel framework tailored for vision graph structures. Our core insight reveals that semantically connected components in the graph exhibit low-rank properties. Building on this observation, we introduce a semantic low-rank prompting method that decomposes low-rank semantic features and integrates them with prompts on vision graph topologies, capturing both global structural patterns and fine-grained semantic dependencies. Extensive experiments demonstrate our method significantly improves ViG’s transfer performance on diverse downstream tasks, achieving results comparable to full fine-tuning while maintaining parameter efficiency.

Training deep neural networks (DNNs) on memory-limited GPUs is challenging, as storing intermediate activations often exceeds available memory. Re-materialization, a technique that preserves exact computations, addresses this by selectively recomputing activations instead of storing them. However, existing methods either fail to scale, lack generality, or introduce excessive execution overhead. We introduce ${\mbox{HiRemate}}$ a ${\textit hierarchical}$ re-materialization framework that recursively partitions large computation graphs, applies optimized solvers at multiple levels, and merges solutions into a global efficient training schedule. This enables scalability to significantly larger graphs than prior ILP-based methods while keeping runtime overhead low. Designed for single-GPU models and activation re-materialization, HiRemate extends the feasibility of training networks with thousands of graph nodes, surpassing prior methods in both efficiency and scalability. Experiments on various types of networks yield up to 50-70% memory reduction with only 10-15% overhead, closely matching optimal solutions while significantly reducing solver time. Seamlessly integrating with PyTorch Autograd, HiRemate requires almost no code change to use, enabling broad adoption in memory-constrained deep learning.