Low-Rank Adaptation

Also known as LoRA

Follow this topic Ask about this topic

Low-rank adaptation studies how to specialize pretrained models by representing weight changes as products of small matrices while keeping most backbone weights frozen. Methods choose where adapters are placed and how their rank is allocated, using approaches such as spectral or gradient-based initialization, adaptive rank selection, and quantized fine-tuning. For multiple tasks or users, adapter routing, expert mixtures, merging, and federated aggregation aim to share useful updates without interference. The central tension is how to gain enough capacity for new tasks while controlling memory and compute, preserving existing capabilities, and generalizing across data shifts; the best rank, placement, and composition strategy depend on the model and task.

Latest papers

542 papers92 papers in the last four weeks, up 59% on the four weeks before. 0.5% of all new papers.

Mar 23, 2026

Parameter-Efficient Fine-Tuning for Medical Text Summarization: A Comparative Study of Lora, Prompt Tuning, and Full Fine-Tuning

Fine-tuning large language models for domain-specific tasks such as medical text summarization demands substantial computational resources. Parameter-efficient fine-tuning (PEFT) methods offer promising alternatives by updating only a small fraction of parameters. This paper compares three adaptation approaches-Low-Rank Adaptation (LoRA), Prompt Tuning, and Full Fine-Tuning-across the Flan-T5 model family on the PubMed medical summarization dataset. Through experiments with multiple random seeds, we demonstrate that LoRA consistently outperforms full fine-tuning, achieving 43.52 +/- 0.18 ROUGE-1 on Flan-T5-Large with only 0.6% trainable parameters compared to 40.67 +/- 0.21 for full fine-tuning. Sensitivity analyses examine the impact of LoRA rank and prompt token count. Our findings suggest the low-rank constraint provides beneficial regularization, challenging assumptions about the necessity of full parameter updates. Code is available at https://github.com/eracoding/llm-medical-summarization

Ulugbek Shernazarov, Rostislav Svitsov, Bin Shi

Telecom SudParis, Institut Polytechnique de Paris, France

Parameter-Efficient Fine-Tuning Text Summarization Low-Rank Adaptation Prompt Tuning

Mar 23, 2026

The Golden Subspace: Where Efficiency Meets Generalization in Continual Test-Time Adaptation

Continual Test-Time Adaptation (CTTA) aims to enable models to adapt online to unlabeled data streams under distribution shift without accessing source data. Existing CTTA methods face an efficiency-generalization trade-off: updating more parameters improves adaptation but severely reduces online inference efficiency. An ideal solution is to achieve comparable adaptation with minimal feature updates; we call this minimal subspace the golden subspace. We prove its existence in a single-step adaptation setting and show that it coincides with the row space of the pretrained classifier. To enable online maintenance of this subspace, we introduce the sample-wise Average Gradient Outer Product (AGOP) as an efficient proxy for estimating the classifier weights without retraining. Building on these insights, we propose Guided Online Low-rank Directional adaptation (GOLD), which uses a lightweight adapter to project features onto the golden subspace and learns a compact scaling vector while the subspace is dynamically updated via AGOP. Extensive experiments on classification and segmentation benchmarks, including autonomous-driving scenarios, demonstrate that GOLD attains superior efficiency, stability, and overall performance. Our code is available at https://github.com/AIGNLAI/GOLD.

Guannan Lai, Da-Wei Zhou, Zhenguo Li +1

School of Artificial Intelligence, Nanjing University · National Key Laboratory for Novel Software Technology, Nanjing University · Hong Kong University of Science and Technology +1

Continual Test-Time Adaptation Low-Rank Adaptation

Mar 23, 2026

Instruction-Tuned, but Not More Verifiable Instruction-Following: A Cross-Task Diagnosis for LoRA Adapters

Adapters are often selected and deployed based on nominal labels (e.g., instruction-tuned), which implicitly suggest what capability improves after adaptation. We test whether nominal training objectives reliably align with realized cross-task capability gains by evaluating the same LoRA adapter across tasks. Our strongest evidence is tied to strict, automatically verifiable instruction following as measured by IFEval: across multiple seeds, base models, and LoRA settings, nominal labels recurrently but not universally fail to predict improvements on this verifiable target, with clear configuration sensitivity including a near-zero or negative case. As an illustrative strongest-case example in a controlled instruction-versus-numeric setting, an instruction-tuned adapter substantially improves off-target NM-based numeric benchmark performance from 0.133 to 0.632 while not improving verifiable instruction following on IFEval (ILA: 0.313 to 0.271; PLA: 0.250 to 0.143; values rounded to three decimals). We refer to this nominal-versus-realized mismatch pattern as capability drift as a descriptive label. The mismatch is visible in the raw cross-task performance matrix; we use a drift score only as a compact summary in the same units as the underlying metrics, not as a new formal metric contribution. Evidence from broader instruction-following benchmarks is benchmark-dependent and mixed, reflecting heterogeneity in how instruction following is operationalized; we therefore do not treat cross-benchmark agreement as a premise. Overall, the practical takeaway is to perform routine cross-task evaluation before deployment and to avoid treating nominal labels as reliable capability proxies.

Junyi Zou

Zjydiary Group

Instruction Tuning Instruction Following Low-Rank Adaptation

Mar 23, 2026

Not All Layers Are Created Equal: Adaptive LoRA Ranks for Personalized Image Generation

Low Rank Adaptation (LoRA) is the de facto fine-tuning strategy to generate personalized images from pre-trained diffusion models. Choosing a good rank is extremely critical, since it trades off performance and memory consumption, but today the decision is often left to the community's consensus, regardless of the personalized subject's complexity. The reason is evident: the cost of selecting a good rank for each LoRA component is combinatorial, so we opt for practical shortcuts such as fixing the same rank for all components. In this paper, we take a first step to overcome this challenge. Inspired by variational methods that learn an adaptive width of neural networks, we let the ranks of each layer freely adapt during fine-tuning on a subject. We achieve it by imposing an ordering of importance on the rank's positions, effectively encouraging the creation of higher ranks when strictly needed. Qualitatively and quantitatively, our approach, LoRA

^2

, achieves a competitive trade-off between DINO, CLIP-I, and CLIP-T across 29 subjects while requiring much less memory and lower rank than high rank LoRA versions. Code: https://github.com/donaldssh/NotAllLayersAreCreatedEqual.

Donald Shenaj, Federico Errica, Antonio Carta

University of Pisa · NEC Laboratories Europe

Low-Rank Adaptation Adaptive Rank Allocation Diffusion Model Adaptation Diffusion Model Fine-Tuning Subject-Driven Image Generation

Mar 22, 2026

Frequency Switching Mechanism for Parameter-E!cient Multi-Task Learning

Multi-task learning (MTL) aims to enable a single model to solve multiple tasks efficiently; however, current parameter-efficient fine-tuning (PEFT) methods remain largely limited to single-task adaptation. We introduce Free Sinewich, a parameter-efficient multi-task learning framework that enables near-zero-cost weight modulation via frequency switching (Free). Specifically, a Sine-AWB (Sinewich) layer combines low-rank factors and convolutional priors into a single kernel, which is then modulated elementwise by a sinusoidal transformation to produce task-specialized weights. A lightweight Clock Net is introduced to produce bounded frequencies that stabilize this modulation during training. Theoretically, sine modulation enhances the rank of low-rank adapters, while frequency separation decorrelates the weights of different tasks. On dense prediction benchmarks, Free Sinewich achieves state-of-the-art performance-efficiency trade-offs (e.g., up to +5.39% improvement over single-task fine-tuning with only 6.53M trainable parameters), offering a compact and scalable paradigm based on frequency-based parameter sharing. Project page: https://casperliuliuliu.github.io/projects/Free-Sinewich.

Shih-Wen Liu, Yen-Chang Chen, Wei-Ta Chu +2

National Cheng Kung University, Tainan, Taiwan · NVIDIA Research, Taiwan

Multi-Task Learning Parameter-Efficient Fine-Tuning Low-Rank Adaptation

Mar 20, 2026

FAAR: Efficient Frequency-Aware Multi-Task Fine-Tuning via Automatic Rank Selection

Adapting models pre-trained on large-scale datasets is a proven way to reach strong performance quickly for down-stream tasks. However, the growth of state-of-the-art mod-els makes traditional full fine-tuning unsuitable and difficult, especially for multi-task learning (MTL) where cost scales with the number of tasks. As a result, recent studies investigate parameter-efficient fine-tuning (PEFT) using low-rank adaptation to significantly reduce the number of trainable parameters. However, these existing methods use a single, fixed rank, which may not be optimal for differ-ent tasks or positions in the MTL architecture. Moreover, these methods fail to learn spatial information that cap-tures inter-task relationships and helps to improve diverse task predictions. This paper introduces Frequency-Aware and Automatic Rank (FAAR) for efficient MTL fine-tuning. Our method introduces Performance-Driven Rank Shrink-ing (PDRS) to allocate the optimal rank per adapter location and per task. Moreover, by analyzing the image frequency spectrum, FAAR proposes a Task-Spectral Pyramidal Decoder (TS-PD) that injects input-specific context into spatial bias learning to better reflect cross-task relationships. Experiments performed on dense visual task benchmarks show the superiority of our method in terms of both accuracy and efficiency compared to other PEFT methods in MTL. FAAR reduces the number of parameters by up to 9 times compared to traditional MTL fine-tuning whilst improving overall performance. Our code is available.

Maxime Fontana, Michael Spratling, Miaojing Shi

King’s College London London, UK · University of Luxembourg Esch-sur-Alzette, Luxembourg · Tongji University Shanghai, China

Nonlinear LoRA Parameter-Efficient Fine-Tuning Multi-Task Learning Low-Rank Adaptation

Mar 18, 2026

Exploring parameter-efficient fine-tuning (PEFT) of billion-parameter vision models with QLoRA and DoRA: insights into generalization for limited-data image classification under a 98:1 test-to-train regime

Automated behavior classification is essential for precision livestock farming but faces challenges of high computational costs and limited labeled data. This study systematically compared three approaches: training from scratch (ResNet-18, ViT-Small), frozen feature extraction, and parameter-efficient fine-tuning (PEFT) of the DINOv3 foundation model (6.7 billion parameters). We evaluated QLoRA and DoRA across multiple configurations varying rank (8, 16, 64) and target modules (q_proj versus all-linear layers). With 2,160 verified training images, we assessed generalization of our model on 211,800 test samples, which is essentially a 98:1 test-to-train ratio. Results demonstrated that PEFT substantially outperformed alternatives, where the best QLoRA configuration (all-linear layers and rank=64) achieved 83.16% test accuracy with only 2.72% parameters (183.0M) in 5.8 hours, compared to 72.87% for ResNet-18 (16.8 hours), 61.91% for ViT-Small (18.7 hours), and 76.56% for frozen DINOv3 (17.5 hours). DoRA achieved comparable accuracy (83.14%) but with longer training time (11.0 hours). Notably, increasing adapter capacity consistently improved generalization while simultaneously not causing overfitting: reducing rank from 16 to 8 decreased test accuracy from 78.38% to 77.17%, while expanding from q_proj-only to all-linear layers with rank=64 improved accuracy from 78.38% to 83.16%. This suggests underfitting, instead of overfitting, is the primary challenge when adapting foundation models to agricultural imagery. Our findings provide guidelines for deploying billion-parameter vision models with PEFT in agricultural livestock applications.

Haiyu Yang, Sumit Sharma, Enhong Liu +1

Cornell University, College of Agriculture and Life Sciences, Ithaca, NY 14853

Precision Livestock Farming Parameter-Efficient Fine-Tuning Vision Foundation Model Adaptation Low-Rank Adaptation Animal Behavior Analysis Quantization-Aware LoRA

Mar 18, 2026

Zipper-LoRA: Dynamic Parameter Decoupling for Speech-LLM based Multilingual Speech Recognition

Speech Large Language Models (Speech-LLMs) have emerged as a powerful approach for automatic speech recognition (ASR) by aligning speech encoders with large language models. However, adapting these systems to multilingual settings with imbalanced data distributions remains challenging. In such scenarios, a stability-plasticity dilemma often arises: fully shared Parameter-Efficient Fine-Tuning (PEFT) can cause negative inter-lingual interference for under-represented languages, while fully language-specific tuning limits the cross-lingual beneficial knowledge transfer needed for low-resource tasks. To address this, we propose Zipper-LoRA, a novel rank-level decoupling framework with three variants (Static, Hard, and Soft) that dynamically synthesizes LoRA updates from shared and language-specific subspaces. By using a lightweight language-conditioned router, Zipper-LoRA dynamically controls the contribution of each subspace at the LoRA rank level, enabling fine-grained sharing where languages are compatible and strict decoupling when conflicts occur. To further stabilize optimization under imbalanced data, we propose a two-stage training strategy with an Initial-B warm start that significantly accelerates convergence. Experiments on a 12-language mixed-resource setting show that Zipper-LoRA consistently outperforms both fully shared and independent baselines, particularly in extremely low-resource scenarios. Moreover, we demonstrate that these gains are robust across both chunked and non-chunked encoder configurations, confirming the framework's reliability for practical, large-scale multilingual ASR. Our code and data will be available at https://github.com/YuCeong-May/Zipper-LoRA for reproducibility.

Yuxiang Mei, Delai Qiu, Shengping Liu +2

Shanghai Engineering Research Center of Intelligent Education and Bigdata, Shanghai Normal University, Shanghai, 200234, China. · Unisound AI Technology Co., Ltd., Beijing, China. · SHNU-Unisound Natural Human-Computer Interaction Lab, Shanghai Normal University.

Multilingual ASR Speech Language Models Parameter-Efficient Fine-Tuning Low-Rank Adaptation Low-Resource Speech Recognition

Mar 18, 2026

Early Intervention for VFM-based Multimodal Medical Image Classification

Current methods for multimodal medical image classification (M3IC) face two major challenges. First, the prevailing "fusion after unimodal image embedding" paradigm cannot fully exploit the complementary and correlated information in the multimodal data. Second, the scarcity of labeled multimodal medical images, coupled with their substantial domain shift from natural images, impedes the use of cutting-edge Vision Foundation Models (VFMs) for medical image embedding. To jointly address the challenges, we propose a novel Early Intervention (EI) framework. Treating one modality as target and the rest as reference, EI harnesses high-level semantic tokens from the reference as intervention tokens to steer the target modality's embedding process at an early stage. For parameter-efficient VFM adaptation, we introduce Mixture of varied-rank LoRAs (MoR), which employs a small set of LoRA experts with distinct ranks and a bypass-allowed router. Extensive experiments on three public datasets, covering retinal disease recognition, skin lesion recognition, and knee anomaly classification, verify the effectiveness of the proposed method against a number of competitive baselines.

Qijie Wei, Hailan Lin, Xirong Li

Renmin University of China, Beijing 100872, China

Cross-Modal Representation Learning Vision Foundation Model Adaptation Low-Rank Adaptation Multimodal Classification Medical Image Classification Multimodal Fusion

Mar 17, 2026

EvoTale: Continual Character Customization for Expanding Story Worlds

Character-centric story visualization aims to synthesize coherent image sequences that depict narrative events and interactions while preserving recurring character identities. In expanding story worlds, new user-specified characters must be continually incorporated despite varying customization difficulty and identity conflicts in multi-character scenes, without disrupting previously learned identities. In this paper, we propose EvoTale, a continual character customization framework for expanding stylized story worlds. We first introduce an All-in-One-World Character Integrator, which accumulates character-specific residual components within a unified LoRA branch using sparsely overlapping subspaces spanned by a shared orthonormal basis, while freezing previously learned components to limit cross-character coupling. We then develop a Character Quality Gate that uses rubric-guided MLLM feedback as a bounded controller to adapt the optimization budget based on the assessed customization quality. Finally, we propose Character-Aware Region-Focus Sampling, which combines bounding-box-guided regional denoising with identity-aware global denoising to preserve character identities within their designated regions while maintaining global narrative coherence. Experimental results show that EvoTale achieves a favorable balance across character fidelity, continual identity retention, multi-character generation quality, and story-text alignment compared with representative story visualization and customization methods.

Jinlu Zhang, Qiyun Wang, Baoxiang Du +6

Key Laboratory of Multimedia Trusted Perception and Efficient Computing, Ministry of Education of China, Xiamen University, 361005, P.R., China. · The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, 510000, P.R., China. · Fuxi AI Lab, NetEase Inc., Hangzhou, 310000, P.R., China.

Image Generation Subject-Driven Image Generation Low-Rank Adaptation

Mar 16, 2026

W2T: LoRA Weights Already Know What They Can Do

Each LoRA checkpoint compactly stores task-specific updates in low-rank weight matrices, offering an efficient way to adapt large language models to new tasks and domains. In principle, these weights already encode what the adapter does and how well it performs. In this paper, we ask whether this information can be read directly from the weights, without running the base model or accessing training data. A key obstacle is that a single LoRA update can be factorized in infinitely many ways. Without resolving this ambiguity, models trained on the factors may fit the particular factorization rather than the underlying update. To this end, we propose \methodfull, which maps each LoRA update to a provably canonical form via QR decomposition followed by SVD, so that all equivalent factorizations share the same representation. The resulting components are then tokenized and processed by a Transformer to produce a weight-space embedding. Across language and vision LoRA collections, W2T achieves strong results on attribute classification, performance prediction, and adapter retrieval, demonstrating that LoRA weights reliably indicate model behavior once factorization ambiguity is removed. Code is available at https://github.com/xiaolonghan2000/Weight2Token.

Xiaolong Han, Ferrante Neri, Zijian Jiang +4

University of Surrey · Stanford University · University of Notre Dame

Weight-Space Learning Low-Rank Adaptation LLM Fine-Tuning

Mar 16, 2026

MoLoRA: Composable Specialization via Per-Token Adapter Routing

Multi-adapter serving systems route entire sequences to a single adapter, forcing a choice when requests span multiple domains. This assumption fails in two important settings: (1) multimodal generation, where text and image tokens require different adapters within the same sequence, and (2) mixed-capability requests like "write code to solve this equation," which need expertise from multiple specialized adapters. We introduce per-token routing, which routes individual tokens to adapters based on either vocabulary structure (for multimodal models) or learned gating (for semantic specialization). Per-token routing is provably optimal, achieving work N for N tokens versus K \cdot N for per-sequence routing with K adapter types. Our key contribution is MoLoRA (Mixture of LoRA), which enables composable specialization: load multiple domain-specific adapters and let a learned router select the appropriate adapter per-token. We demonstrate that specialization dramatically beats scale: MoLoRA enables Qwen3-1.7B to exceed Qwen3-8B across four reasoning benchmarks while being 4.7x smaller. This enables modular expertise at inference time: train focused LoRAs independently, combine them without retraining, and add new capabilities by simply loading new adapters.

Shrey Shah, Justin Wagle

LLM Routing Low-Rank Adaptation

Mar 16, 2026

AdapterTune: Zero-Initialized Low-Rank Adapters for Frozen Vision Transformers

Frozen-backbone transfer with Vision Transformers faces two under-addressed issues: optimization instability when adapters are naively inserted into a fixed feature extractor, and the absence of principled guidance for setting adapter capacity. We introduce AdapterTune, which augments each transformer block with a residual low-rank bottleneck whose up-projection is zero-initialized, guaranteeing that the adapted network starts exactly at the pretrained function and eliminates early-epoch representation drift. On the analytical side, we formalize adapter rank as a capacity budget for approximating downstream task shifts in feature space. The resulting excess-risk decomposition predicts monotonic but diminishing accuracy gains with increasing rank, an ``elbow'' behavior we confirm through controlled sweeps. We evaluate on 9 datasets and 3 backbone scales with multi-seed reporting throughout. On a core 5 dataset transfer suite, AdapterTune improves top-1 accuracy over head-only transfer by +14.9 points on average while training only 0.92 of the parameters required by full fine-tuning, and outperforms full fine-tuning on 10 of 15 dataset-backbone pairs. Across the full benchmark, AdapterTune improves over head-only transfer on every dataset-backbone pair tested. Ablations on rank, placement, and initialization isolate each design choice. The code is available at: https://github.com/salimkhazem/adaptertune

Salim Khazem

Talan Research Center, Paris, France

Adapter Tuning Low-Rank Adaptation Parameter-Efficient Fine-Tuning Vision Transformer

Mar 15, 2026

Not All Directions Matter: Towards Structured and Task-Aware Low-Rank Model Adaptation

Low-Rank Adaptation (LoRA) has become a cornerstone of parameter-efficient fine-tuning (PEFT). Yet, its efficacy is hampered by two fundamental limitations: semantic drift, by treating all update directions with equal importance, and structural incoherence, from adapting layers independently, resulting in suboptimal, uncoordinated updates. To remedy these, we propose StructLoRA, a framework that addresses both limitations through a principled, dual-component design: (1) an Information Bottleneck-guided filter that prunes task-irrelevant directions to mitigate semantic drift, and (2) a lightweight, training-only graph-based coordinator that enforces inter-layer consistency to resolve structural incoherence. Extensive experiments across large language model , vision language model, and vision model (including LLaMA, LLaVA, and ViT) demonstrate that StructLoRA consistently establishes a new state-of-the-art, outperforming not only vanilla LoRA but also advanced dynamic rank allocation and sparsity-based methods. Notably, the benefits are particularly pronounced in challenging low-rank and low-data regimes. Crucially, since our proposed modules operate only during training, StructLoRA enhances performance with zero additional inference cost, advancing the focus of PEFT -- from mere parameter compression to a more holistic optimization of information quality and structural integrity.

Xi Xiao, Chenrui Ma, Yunbei Zhang +7

University of Alabama at Birmingham · University of Virginia · Tulane University +5

Low-Rank Adaptation Parameter-Efficient Fine-Tuning

Mar 14, 2026

IGU-LoRA: Adaptive Rank Allocation via Integrated Gradients and Uncertainty-Aware Scoring

As large language models (LLMs) scale to billions of parameters, full-parameter fine-tuning becomes compute- and memory-prohibitive. Parameter-efficient fine-tuning (PEFT) mitigates this issue by updating only a small set of task-specific parameters while keeping the base model frozen. Among PEFT approaches, low-rank adaptation (LoRA) is widely adopted; however, it enforces a uniform rank across layers despite substantial variation in layer importance, motivating {layerwise} rank allocation. Recent adaptive-rank variants (e.g., AdaLoRA) allocate ranks based on importance scores, yet typically rely on instantaneous gradients that capture only local sensitivity, overlooking non-local, pathwise effects within the same layer, which yields unstable and biased scores. To address this limitation, we introduce IGU-LoRA, an adaptive-rank LoRA that (i) computes within-layer Integrated Gradients (IG) sensitivities and aggregates them into a layer-level score for rank allocation, and (ii) applies an uncertainty-aware scheme using exponential moving averages with deviation tracking to suppress noisy updates and calibrate rank selection. Theoretically, we prove an upper bound on the composite trapezoidal rule approximation error for parameter-space IG under a pathwise Hessian-Lipschitz condition, which informs the quadrature budget. Across diverse tasks and architectures, IGU-LoRA consistently outperforms strong PEFT baselines at matched parameter budgets, improving downstream accuracy and robustness. Ablations confirm the contributions of pathwise within-layer sensitivity estimates and uncertainty-aware selection to effective rank allocation. Our code is publicly available at https://github.com/withyou12/igulora.git

Xuan Cui, Huiyue Li, Run Zeng +5

Chongqing Technology and Business University · Shanghai Jiao Tong University

LLM Fine-Tuning Low-Rank Adaptation Adaptive Rank Allocation Parameter-Efficient Fine-Tuning

Mar 13, 2026

Expert Pyramid Tuning: Efficient Parameter Fine-Tuning for Expertise-Driven Task Allocation

Parameter-Efficient Fine-Tuning (PEFT) has become a dominant paradigm for deploying LLMs in multi-task scenarios due to its extreme parameter efficiency. While Mixture-of-Experts (MoE) based LoRA variants have achieved promising results by dynamically routing tokens to different low-rank experts, they largely overlook the hierarchical nature of task complexity. Existing methods typically employ experts with uniform architectures, limiting their ability to capture diverse feature granularities required by distinct tasks--where some tasks demand high-level semantic abstraction while others require fine-grained syntactic manipulation. To bridge this gap, we propose Expert Pyramid Tuning (EPT), a novel architecture that integrates the multi-scale feature pyramid concept from computer vision into the realm of PEFT. Unlike standard LoRA, EPT decomposes task adaptation into two stages: (1) A shared meta-knowledge Subspace that encodes universal linguistic patterns in low dimensions; (2) A Pyramid Projection Mechanism that utilizes learnable up-projection operators to reconstruct high-dimensional features at varying scales. A task-aware router then dynamically selects the optimal combination of these multi-scale features. Extensive experiments across multiple multi-task benchmarks demonstrate that EPT significantly outperforms SOTA MoE-LoRA variants. Crucially, thanks to the re-parameterization capability of our design, EPT achieves this performance improvement while simultaneously reducing the number of training parameters.

Jia-Chen Zhang, Zhen-Wei Yan, Yu-Jie Xiong +1

School of Electronic and Electrical Engineering, Shanghai University of Engineering Science 333 Longteng Road, Shanghai, China

Parameter-Efficient Fine-Tuning Low-Rank Adaptation Mixture-of-Experts Models Multi-Task Learning

Mar 12, 2026

NeuroLoRA: Context-Aware Neuromodulation for Parameter-Efficient Multi-Task Adaptation

Parameter-Efficient Fine-Tuning (PEFT) techniques, particularly Low-Rank Adaptation (LoRA), have become essential for adapting Large Language Models (LLMs) to downstream tasks. While the recent FlyLoRA framework successfully leverages bio-inspired sparse random projections to mitigate parameter interference, it relies on a static, magnitude-based routing mechanism that is agnostic to input context. In this paper, we propose NeuroLoRA, a novel Mixture-of-Experts (MoE) based LoRA framework inspired by biological neuromodulation -- the dynamic regulation of neuronal excitability based on context. NeuroLoRA retains the computational efficiency of frozen random projections while introducing a lightweight, learnable neuromodulation gate that contextually rescales the projection space prior to expert selection. We further propose a Contrastive Orthogonality Loss to explicitly enforce separation between expert subspaces, enhancing both task decoupling and continual learning capacity. Extensive experiments on MMLU, GSM8K, and ScienceQA demonstrate that NeuroLoRA consistently outperforms FlyLoRA and other strong baselines across single-task adaptation, multi-task model merging, and sequential continual learning scenarios, while maintaining comparable parameter efficiency.

Yuxin Yang, Haoran Zhang, Mingxuan Li +6

Shanghai University · Fudan University

Parameter-Efficient Fine-Tuning Low-Rank Adaptation Multi-Task Learning Continual Learning for LLMs

Mar 12, 2026

AdaFuse: Accelerating Dynamic Adapter Inference via Token-Level Pre-Gating and Fused Kernel Optimization

The integration of dynamic, sparse structures like Mixture-of-Experts (MoE) with parameter-efficient adapters (e.g., LoRA) is a powerful technique for enhancing Large Language Models (LLMs). However, this architectural enhancement comes at a steep cost: despite minimal increases in computational load, the inference latency often skyrockets, leading to decoding speeds slowing by over 2.5 times. Through a fine-grained performance analysis, we pinpoint the primary bottleneck not in the computation itself, but in the severe overhead from fragmented, sequential CUDA kernel launches required for conventional dynamic routing. To address this challenge, we introduce AdaFuse, a framework built on a tight co-design between the algorithm and the underlying hardware system to enable efficient dynamic adapter execution. Departing from conventional layer-wise or block-wise routing, AdaFuse employs a token-level pre-gating strategy, which makes a single, global routing decision for all adapter layers before a token is processed. This "decide-once, apply-everywhere" approach effectively staticizes the execution path for each token, creating an opportunity for holistic optimization. We capitalize on this by developing a custom CUDA kernel that performs a fused switching operation, merging the parameters of all selected LoRA adapters into the backbone model in a single, efficient pass. Experimental results on popular open-source LLMs show that AdaFuse achieves accuracy on par with state-of-the-art dynamic adapters while drastically cutting decoding latency by a factor of over 2.4x, thereby bridging the gap between model capability and inference efficiency.

Qiyang Li, Rui Kong, Yuchen Li +5

Baidu Inc. · Shanghai Jiao Tong University

LLM Inference Acceleration Low-Rank Adaptation

Mar 11, 2026

Reversible Lifelong Model Editing via Semantic Routing-Based LoRA

The dynamic evolution of real-world necessitates model editing within Large Language Models. While existing methods explore modular isolation or parameter-efficient strategies, they still suffer from semantic drift or knowledge forgetting due to continual updating. To address these challenges, we propose SoLA, a Semantic routing-based LoRA framework for lifelong model editing. In SoLA, each edit is encapsulated as an independent LoRA module, which is frozen after training and mapped to input by semantic routing, allowing dynamic activation of LoRA modules via semantic matching. This mechanism avoids semantic drift caused by cluster updating and mitigates catastrophic forgetting from parameter sharing. More importantly, SoLA supports precise revocation of specific edits by removing key from semantic routing, which restores model's original behavior. To our knowledge, this reversible rollback editing capability is the first to be achieved in existing literature. Furthermore, SoLA integrates decision-making process into edited layer, eliminating the need for auxiliary routing networks and enabling end-to-end decision-making process. Extensive experiments demonstrate that SoLA effectively learns and retains edited knowledge, achieving accurate, efficient, and reversible lifelong model editing.

Haihua Luo, Xuming Ran, Tommi Kärkkäinen +4

Faculty of Information Technology, University of Jyväskylä, Jyväskylä, Finland · National University of Singapore, Singapore · School of Computer Science and Technology, Dalian University of Technology, Dalian, China

Model Editing Language Model Editing Continual Learning for LLMs Low-Rank Adaptation

Mar 11, 2026

Med-DualLoRA: Local Adaptation of Foundation Models for 3D Cardiac MRI

Foundation models (FMs) show great promise for robust downstream performance across medical imaging tasks and modalities, including cardiac magnetic resonance (CMR), following task-specific adaptation. However, adaptation using single-site data may lead to suboptimal performance and increased model bias, while centralized fine-tuning on clinical data is often infeasible due to privacy constraints. Federated fine-tuning offers a privacy-preserving alternative; yet conventional approaches struggle under heterogeneous, non-IID multi-center data and incur substantial communication overhead when adapting large models. In this work, we study federated FM fine-tuning for 3D CMR disease detection and propose Med-DualLoRA, a client-aware parameter-efficient fine-tuning (PEFT) federated framework that disentangles globally shared and local low-rank adaptations (LoRA) through additive decomposition. Global and local LoRA modules are trained locally, but only the global component is shared and aggregated across sites, keeping local adapters private. This design improves personalization while significantly reducing communication cost, and experiments show that adapting only two transformer blocks preserves performance while further improving efficiency. We evaluate our method on a multi-center state-of-the-art cine 3D CMR FM fine-tuned for disease detection using ACDC and combined M&Ms datasets, treating each vendor as a federated client. Med-DualLoRA achieves statistically significant improved performance (balanced accuracy 0.768, specificity 0.612) compared to other federated PEFT baselines, while maintaining communication efficiency. Our approach provides a scalable solution for local federated adaptation of medical FMs under realistic clinical constraints.

Joan Perramon-Llussà, Amelia Jiménez-Sánchez, Grzegorz Skorupko +4

Artificial Intelligence in Medicine Lab (BCN-AIM), Departament de Matemàtiques i Informàtica, Universitat de Barcelona, Spain · Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain

Federated Fine-Tuning Medical Imaging Foundation Models Federated Learning Low-Rank Adaptation Cardiac MRI

Mar 11, 2026

Bilevel Layer-Positioning LoRA for Real Image Dehazing

Learning-based real image dehazing methods have achieved notable progress, yet they still face adaptation challenges in diverse real haze scenes. These challenges mainly stem from the lack of effective unsupervised mechanisms for unlabeled data and the heavy cost of full model fine-tuning. To address these challenges, we propose the haze-to-clear text-directed loss that leverages CLIP's cross-modal capabilities to reformulate real image dehazing as a semantic alignment problem in latent space, thereby providing explicit unsupervised cross-modal guidance in the absence of reference images. Furthermore, we introduce the Bilevel Layer-positioning LoRA (BiLaLoRA) strategy, which learns both the LoRA parameters and automatically search the injection layers, enabling targeted adaptation of critical network layers. Extensive experiments demonstrate our superiority against state-of-the-art methods on multiple real-world dehazing benchmarks. The code is publicly available at https://github.com/YanZhang-zy/BiLaLoRA.

Yan Zhang, Long Ma, Yuxin Feng +3

School of Computer Science and Engineering, Sun Yat-sen University · Dalian University of Technology · Xidian University +1

Image Dehazing Low-Rank Adaptation

Mar 10, 2026

ID-LoRA: Identity-Driven Audio-Video Personalization with In-Context LoRA

Existing video personalization methods preserve visual likeness but treat video and audio separately. Without access to the visual scene, audio models cannot synchronize sounds with on-screen actions; and because classical voice-cloning models condition only on a reference recording, a text prompt cannot redirect speaking style or acoustic environment. We propose ID-LoRA (Identity-Driven In-Context LoRA), which jointly generates a subject's appearance and voice in a single model, letting a text prompt, a reference image, and a short audio clip govern both modalities together. ID-LoRA adapts the LTX-2 joint audio-video diffusion backbone via parameter-efficient In-Context LoRA and, to our knowledge, is the first method to personalize visual appearance and voice in a single generative pass. Two challenges arise. Reference and generation tokens share the same positional-encoding space, making them hard to distinguish; we address this with negative temporal positions, placing reference tokens in a disjoint RoPE region while preserving their internal temporal structure. Speaker characteristics also tend to be diluted during denoising; we introduce identity guidance, a classifier-free guidance variant that amplifies speaker-specific features by contrasting predictions with and without the reference signal. In human preference studies, ID-LoRA is preferred over Kling 2.6 Pro by 73% of annotators for voice similarity and 65% for speaking style. On cross-environment settings, speaker similarity improves by 24% over Kling, with the gap widening as conditions diverge. A preliminary user study further suggests that joint generation provides a useful inductive bias for physically grounded sound synthesis. ID-LoRA achieves these results with only ~3K training pairs on a single GPU. Code, models, and data will be released.

Aviad Dahan, Moran Yanuka, Noa Kraicer +2

Tel Aviv University

Audio-Video Generation Identity-Preserving Video Generation Voice Cloning Low-Rank Adaptation

Mar 10, 2026

ReMix: Reinforcement routing for mixtures of LoRAs in LLM finetuning

Low-rank adapters (LoRAs) are a parameter-efficient finetuning technique that injects trainable low-rank matrices into pretrained models to adapt them to new tasks. Mixture-of-LoRAs models expand neural networks efficiently by routing each layer input to a small subset of specialized LoRAs of the layer. Existing Mixture-of-LoRAs routers assign a learned routing weight to each LoRA to enable end-to-end training of the router. Despite their empirical promise, we observe that the routing weights are typically extremely imbalanced across LoRAs in practice, where only one or two LoRAs often dominate the routing weights. This essentially limits the number of effective LoRAs and thus severely hinders the expressive power of existing Mixture-of-LoRAs models. In this work, we attribute this weakness to the nature of learnable routing weights and rethink the fundamental design of the router. To address this critical issue, we propose a new router designed that we call Reinforcement Routing for Mixture-of-LoRAs (ReMix). Our key idea is using non-learnable routing weights to ensure all active LoRAs to be equally effective, with no LoRA dominating the routing weights. However, our routers cannot be trained directly via gradient descent due to our non-learnable routing weights. Hence, we further propose an unbiased gradient estimator for the router by employing the reinforce leave-one-out (RLOO) technique, where we regard the supervision loss as the reward and the router as the policy in reinforcement learning. Our gradient estimator also enables to scale up training compute to boost the predictive performance of our ReMix. Extensive experiments demonstrate that our proposed ReMix significantly outperform state-of-the-art parameter-efficient finetuning methods under a comparable number of activated parameters.

Ruizhong Qiu, Hanqing Zeng, Yinglong Xia +15

University of Illinois Urbana-Champaign, IL, USA · Meta AI Menlo Park, CA, USA · Washington University St. Louis, WA, USA +1

Parameter-Efficient Fine-Tuning Low-Rank Adaptation

Mar 10, 2026

TemporalDoRA: Temporal PEFT for Robust Surgical Video Question Answering

Surgical Video Question Answering (VideoQA) requires accurate temporal grounding while remaining robust to natural variation in how clinicians phrase questions, where linguistic bias can arise. Standard Parameter Efficient Fine Tuning (PEFT) methods adapt pretrained projections without explicitly modeling frame-to-frame interactions within the adaptation pathway, limiting their ability to exploit sparse temporal evidence. We introduce TemporalDoRA, a video-specific PEFT formulation that extends Weight-Decomposed Low-Rank Adaptation by (i) inserting lightweight temporal Multi-Head Attention (MHA) inside the low-rank bottleneck of the vision encoder and (ii) selectively applying weight decomposition only to the trainable low-rank branch rather than the full adapted weight. This design enables temporally-aware updates while preserving a frozen backbone and stable scaling. By mixing information across frames within the adaptation subspace, TemporalDoRA steers updates toward temporally consistent visual cues and improves robustness with minimal parameter overhead. To benchmark this setting, we present REAL-Colon-VQA, a colonoscopy VideoQA dataset with 6,424 clip--question pairs, including paired rephrased Out-of-Template questions to evaluate sensitivity to linguistic variation. TemporalDoRA improves Out-of-Template performance, and ablation studies confirm that temporal mixing inside the low-rank branch is the primary driver of these gains. We also validate on EndoVis18-VQA adapted to short clips and observe consistent improvements on the Out-of-Template split. Code and dataset available at~Anonymous GitHub.

Luca Carlini, Chiara Lena, Cesare Hassan +4

Dipartimento di Elettronica, Informazione e Bioingegneria (DEIB), Politecnico di Milano, Italy · IRCCS Humanitas Research Hospital, Italy · UCL Hawkes Institute and Department of Computer Science, University College London, UK +1

Video QA Parameter-Efficient Fine-Tuning Medical VQA Low-Rank Adaptation

Mar 10, 2026

On Catastrophic Forgetting in Low-Rank Decomposition-Based Parameter-Efficient Fine-Tuning

Parameter-efficient fine-tuning (PEFT) based on low-rank decomposition, such as LoRA, has become a standard for adapting large pretrained models. However, its behavior in sequential learning -- specifically regarding catastrophic forgetting -- remains insufficiently understood. In this work, we present an empirical study showing that forgetting is strongly influenced by the geometry and parameterization of the update subspace. While methods that restrict updates to small, shared matrix subspaces often suffer from task interference, tensor-based decompositions (e.g., LoRETTA) mitigate forgetting by capturing richer structural information within ultra-compact budgets, and structurally aligned parameterizations (e.g., WeGeFT) preserve pretrained representations. Our findings highlight update subspace design as a key factor in continual learning and offer practical guidance for selecting efficient adaptation strategies in sequential settings.

Muhammad Ahmad, Jingjing Zheng, Yankai Cao

Department of Electrical and Computer Engineering, The University of British Columbia · Department of Mathematics, The University of British Columbia · Department of Chemical and Biological Engineering, The University of British Columbia

Parameter-Efficient Fine-Tuning Catastrophic Forgetting Low-Rank Adaptation Continual Learning for LLMs

Mar 10, 2026

ForgeDreamer: Industrial Text-to-3D Generation with Multi-Expert LoRA and Cross-View Hypergraph

Current text-to-3D generation methods excel in natural scenes but struggle with industrial applications due to two critical limitations: domain adaptation challenges where conventional LoRA fusion causes knowledge interference across categories, and geometric reasoning deficiencies where pairwise consistency constraints fail to capture higher-order structural dependencies essential for precision manufacturing. We propose a novel framework named ForgeDreamer addressing both challenges through two key innovations. First, we introduce a Multi-Expert LoRA Ensemble mechanism that consolidates multiple category-specific LoRA models into a unified representation, achieving superior cross-category generalization while eliminating knowledge interference. Second, building on enhanced semantic understanding, we develop a Cross-View Hypergraph Geometric Enhancement approach that captures structural dependencies spanning multiple viewpoints simultaneously. These components work synergistically improved semantic understanding, enables more effective geometric reasoning, while hypergraph modeling ensures manufacturing-level consistency. Extensive experiments on a custom industrial dataset demonstrate superior semantic generalization and enhanced geometric fidelity compared to state-of-the-art approaches. Code is available at https://github.com/Junhaocai27/ForgeDreamer

Junhao Cai, Deyu Zeng, Junhao Pang +3

Shenzhen University, Shenzhen, Guangdong 518060, China · Guangzhou Maritime University, Guangzhou, Guangdong 510725, China

Text-to-3D Generation Low-Rank Adaptation

Mar 9, 2026

Alfa: Attentive Low-Rank Filter Adaptation for Structure-Aware Cross-Domain Personalized Gaze Estimation

Pre-trained gaze models learn to identify useful patterns commonly found across users, but subtle user-specific variations (i.e., eyelid shape or facial structure) can degrade model performance. Test-time personalization (TTP) adapts pre-trained models to these user-specific domain shifts using only a few unlabeled samples. Efficient fine-tuning is critical in performing this domain adaptation: data and computation resources can be limited-especially for on-device customization. While popular parameter-efficient fine-tuning (PEFT) methods address adaptation costs by updating only a small set of weights, they may not be taking full advantage of structures encoded in pre-trained filters. To more effectively leverage existing structures learned during pre-training, we reframe personalization as a process to reweight existing features rather than learning entirely new ones. We present Attentive Low-Rank Filter Adaptation (Alfa) to adapt gaze models by reweighting semantic patterns in pre-trained filters. With Alfa, singular value decomposition (SVD) extracts dominant spatial components that capture eye and facial characteristics across users. Via an attention mechanism, we need only a few unlabeled samples to adjust and reweight pre-trained structures, selectively amplifying those relevant to a target user. Alfa achieves the lowest average gaze errors across four cross-dataset gaze benchmarks, outperforming existing TTP methods and low-rank adaptation (LoRA)-based variants. We also show that Alfa's attentive low-rank methods can be applied to applications beyond vision, such as diffusion-based language models.

He-Yen Hsieh, Wei-Te Mark Ting, H. T. Kung

Harvard University

Gaze Estimation Low-Rank Adaptation

Mar 9, 2026

Video2LoRA: Unified Semantic-Controlled Video Generation via Per-Reference-Video LoRA

Achieving semantic alignment across diverse video generation conditions remains a significant challenge. Methods that rely on explicit structural guidance often enforce rigid spatial constraints that limit semantic flexibility, whereas models tailored for individual control types lack interoperability and adaptability. These design bottlenecks hinder progress toward flexible and efficient semantic video generation. To address this, we propose Video2LoRA, a scalable and generalizable framework for semantic-controlled video generation that conditions on a reference video. Video2LoRA employs a lightweight hypernetwork to predict personalized LoRA weights for each semantic input, which are combined with auxiliary matrices to form adaptive LoRA modules integrated into a frozen diffusion backbone. This design enables the model to generate videos consistent with the reference semantics while preserving key style and content variations, eliminating the need for any per-condition training. Notably, the final model weights less than 150MB, making it highly efficient for storage and deployment. Video2LoRA achieves coherent, semantically aligned generation across diverse conditions and exhibits strong zero-shot generalization to unseen semantics.

Zexi Wu, Baolu Li, Jing Dai +5

Dalian University of Technology · HKUST

Controllable Video Generation Video-Reference Video Generation Video Diffusion Models Low-Rank Adaptation

Mar 9, 2026

Stabilized Fine-Tuning with LoRA in Federated Learning: Mitigating the Side Effect of Client Size and Rank via the Scaling Factor

Large Language Models (LLMs) are pivotal in natural language processing. The impracticality of full fine-tuning has prompted Parameter-Efficient Fine-Tuning (PEFT) methods like Low-Rank Adaptation (LoRA), optimizing low-rank matrices A and B. In distributed scenarios where privacy constraints necessitate Federated Learning (FL), however, the integration of LoRA is often unstable. Specifically, we identify that aggregating updates from multiple clients introduces statistical variance that scales with the client count, causing gradient collapse when using high-rank adapters. Existing scaling factor candidates, such as the one used by Rank-Stabilized LoRA, ignore the interaction caused by the aggregation process. To bridge this gap, this paper introduces Stabilized Federated LoRA (SFed-LoRA), a framework that theoretically characterizes the interaction between adapter rank and federated aggregation. We derive an optimal scaling factor designed to effectively mitigate the aggregation error accumulating across N clients. By correcting the scaling mismatch inherent in previous approaches, SFed-LoRA restores the efficacy of high-rank adaptation without altering the original model architecture or increasing inference latency. Extensive experiments in diverse tasks, model architectures, and heterogeneous data distributions are conducted to validate our results. We demonstrate that SFed-LoRA prevents high-rank collapse, and achieves significantly improved stability and faster convergence compared with state-of-the-art baselines for high-rank adaptation.

Jiayu Huang, Xiaohu Wu, Tiantian He +1

Beijing University of Posts and Telecommunications, Beijing, China · Agency for Science, Technology and Research, Singapore, Singapore

Federated Fine-Tuning Low-Rank Adaptation Federated Learning Federated Learning Aggregation

Mar 9, 2026

FedMomentum: Preserving LoRA Training Momentum in Federated Fine-Tuning

Federated fine-tuning of large language models (LLMs) with low-rank adaptation (LoRA) offers a communication-efficient and privacy-preserving solution for task-specific adaptation. Naive aggregation of LoRA modules introduces noise due to mathematical incorrectness when averaging the downsampling and upsampling matrices independently. However, existing noise-free aggregation strategies inevitably compromise the structural expressiveness of LoRA, limiting its ability to retain client-specific adaptations by either improperly reconstructing the low-rank structure or excluding partially trainable components. We identify this problem as loss of training momentum, where LoRA updates fail to accumulate effectively across rounds, resulting in slower convergence and suboptimal performance. To address this, we propose FedMomentum, a novel framework that enables structured and momentum-preserving LoRA aggregation via singular value decomposition (SVD). Specifically, after aggregating low-rank updates in a mathematically correct manner, FedMomentum applies SVD to extract the dominant components that capture the main update directions. These components are used to reconstruct the LoRA modules with the same rank, while residual components can be retained and later merged into the backbone to preserve semantic information and ensure robustness. Extensive experiments across multiple tasks demonstrate that FedMomentum consistently outperforms prior state-of-the-art methods in convergence speed and final accuracy.

Peishen Yan, Yang Hua, Hao Wang +4

Shanghai Jiao Tong University · Queen’s University Belfast · Stevens Institute of Technology +2

Federated Fine-Tuning Federated Learning Aggregation Federated Learning Low-Rank Adaptation