Fine-tuning large language models (LLMs) is crucial for adapting them to specific tasks, yet it remains computationally demanding and raises concerns about correctness and privacy, particularly in untrusted environments. Although parameter-efficient methods like Low-Rank Adaptation (LoRA) significantly reduce resource requirements, ensuring the security and verifiability of fine-tuning under zero-knowledge constraints remains an unresolved challenge. To address this, we introduce VeriLoRA, the first framework to integrate LoRA fine-tuning with zero-knowledge proofs (ZKPs), achieving provable security and correctness. VeriLoRA employs advanced cryptographic techniques -- such as lookup arguments, sumcheck protocols, and polynomial commitments -- to verify both arithmetic and non-arithmetic operations in Transformer-based architectures. The framework provides end-to-end verifiability for forward propagation, backward propagation, and parameter updates during LoRA fine-tuning, while safeguarding the privacy of model parameters and training data. Leveraging GPU-based implementations, VeriLoRA demonstrates practicality and efficiency through experimental validation on open-source LLMs like LLaMA, scaling up to 13 billion parameters. By combining parameter-efficient fine-tuning with ZKPs, VeriLoRA bridges a critical gap, enabling secure and trustworthy deployment of LLMs in sensitive or untrusted environments.

Low-rank approximation methods such as singular value decomposition (SVD) and its variants (e.g., Fisher-weighted SVD, Activation SVD) have recently emerged as effective tools for neural network compression. In this setting, decomposition acts as a "surgical" intervention, followed by fine-tuning that serves as "rehab" to recover accuracy. Inspired by prehabilitation in surgery, we introduce a pre-compression fine-tuning stage, Low-Rank Prehab, that explicitly encourages low-rank structure in weight matrices while preserving task performance. By conditioning the model before SVD, Prehab steers weights toward spectrally compact regions of the parameter space, enabling smoother low-rank approximation and improved recovery. Experiments on large language models (LLMs) and other Transformer-based architectures, including Vision Transformers (ViTs), show that Prehab substantially reduces the immediate accuracy drop after compression and consistently improves post-finetuning performance. Across a wide range of compression ratios, our method outperforms state-of-the-art SVD-based techniques such as SVD-LLM, highlighting the importance of preparing models for compression rather than only improving the compression and recovery stages. Source code is available at https://github.com/niqretnuh/PREHAB-SVD

A challenge in fine-tuning text-to-image diffusion models for specific topics is to select good examples. Fine-tuning from image sets of varying quality, such as Wikipedia Commons, will often produce poor output. However, training images that \textit{do} exemplify the target concept (e.g., a \textit{female Mountain Bluebird}) help ensure that the generated images are similarly representative (e.g., have the prototypical blue-wings and gray chest). In this work, we propose QZLoRA, a framework to select images for low-rank adaptation (LoRA). The approach leverages QuizRank, a method to automatically rank images by treating them as an `educational intervention' and `quizzing' a VLM. We demonstrate that QZLoRA can produce better aligned, photorealistic images with fewer samples. We also show that these fine-tuned models can produce stylized that are similarly representative (i.e., illustrations). Our results highlight the promise of combining automated visual reasoning with parameter-efficient fine-tuning for topic-adaptive generative modeling.

Efficient large-scale retrieval requires representations that are both compact and discriminative. Foundation models provide powerful visual and multimodal embeddings, but nearest neighbor search in these high-dimensional spaces is computationally expensive. Hashing offers an efficient alternative by enabling fast Hamming distance search with binary codes, yet existing approaches often rely on complex pipelines, multi-term objectives, designs specialized for a single learning paradigm, and long training times. We introduce CroVCA (Cross-View Code Alignment), a simple and unified principle for learning binary codes that remain consistent across semantically aligned views. A single binary cross-entropy loss enforces alignment, while coding-rate maximization serves as an anti-collapse regularizer to promote balanced and diverse codes. To implement this, we design HashCoder, a lightweight MLP hashing network with a final batch normalization layer to enforce balanced codes. HashCoder can be used as a probing head on frozen embeddings or to adapt encoders efficiently via LoRA fine-tuning. Across benchmarks, CroVCA achieves state-of-the-art results in just 5 training epochs. At 16 bits, it particularly well-for instance, unsupervised hashing on COCO completes in under 2 minutes and supervised hashing on ImageNet100 in about 3 minutes on a single GPU. These results highlight CroVCA's efficiency, adaptability, and broad applicability.

Low-bit floating-point (FP) formats, such as FP8, provide significant acceleration and memory savings in model training thanks to native hardware support on modern GPUs and NPUs. However, we analyze that FP8 quantization offers speedup primarily for large-dimensional matrix multiplications, while inherent quantization overheads diminish speedup when applied to low-rank adaptation (LoRA), which uses small-dimensional matrices for efficient fine-tuning of large language models (LLMs). To address this limitation, we propose FALQON, a novel framework that eliminates the quantization overhead from separate LoRA computational paths by directly merging LoRA adapters into an FP8-quantized backbone during fine-tuning. Furthermore, we reformulate the forward and backward computations for merged adapters to significantly reduce quantization overhead, and introduce a row-wise proxy update mechanism that efficiently integrates substantial updates into the quantized backbone. Experimental evaluations demonstrate that FALQON achieves approximately a 3$\times$ training speedup over existing quantized LoRA methods with a similar level of accuracy, providing a practical solution for efficient large-scale model fine-tuning. Moreover, FALQON's end-to-end FP8 workflow removes the need for post-training quantization, facilitating efficient deployment. Code is available at https://github.com/iamkanghyunchoi/falqon.

Recent work in benchmarking bias and fairness in speech large language models (SpeechLLMs) has relied heavily on multiple-choice question answering (MCQA) formats. The model is tasked to choose between stereotypical, anti-stereotypical, or neutral/irrelevant answers given an input speech prompt and an optional text prompt. Such MCQA benchmarks implicitly assume that model performance is consistent across other MCQA tasks, voices, and other task formats such as more realistic, long-form evaluations. In this paper, we probe that assumption. We fine-tune three SpeechLLMs using LoRA adapters to induce specific MCQA behaviours: preference for stereotypical, anti-stereotypical, or neutral/uncertain answers. We then evaluate whether these behaviours generalise to another, distinct MCQA benchmark, and more critically to long-form, creative generation tasks. Our results show that performance on MCQA bias benchmarks fails to reliably predict performances across other MCQA benchmarks, and more importantly across long-form tasks. We conclude that current MCQA bias benchmarks show limited evidence of cross-task generalisation in the speech domain, and also propose an evaluation suite for measuring behaviour transferability in future models and benchmarks.

We conduct a behavioral exploration of LoRA fine-tuned LLMs for Passage Reranking to understand how relevance signals are learned and deployed by Large Language Models. By fine-tuning Mistral-7B, LLaMA3.1-8B, and Pythia-6.9B on MS MARCO under diverse LoRA configurations, we investigate how relevance modeling evolves across checkpoints, the impact of LoRA rank (1, 2, 8, 32), and the relative importance of updated MHA vs. MLP components. Our ablations reveal which layers and projections within LoRA transformations are most critical for reranking accuracy. These findings offer fresh explanations into LoRA's adaptation mechanisms, setting the stage for deeper mechanistic studies in Information Retrieval. All models used in this study have been shared.

Language Models (LMs) typically adhere to a "pre-training and fine-tuning" paradigm, where a universal pre-trained model can be fine-tuned to cater to various specialized domains. Low-Rank Adaptation (LoRA) has gained the most widespread use in LM fine-tuning due to its lightweight computational cost and remarkable performance. Because the proportion of parameters tuned by LoRA is relatively small, there might be a misleading impression that the LoRA fine-tuning data is invulnerable to Membership Inference Attacks (MIAs). However, we identify that utilizing the pre-trained model can induce more information leakage, which is neglected by existing MIAs. Therefore, we introduce LoRA-Leak, a holistic evaluation framework for MIAs against the fine-tuning datasets of LMs. LoRA-Leak incorporates fifteen membership inference attacks, including ten existing MIAs, and five improved MIAs that leverage the pre-trained model as a reference. In experiments, we apply LoRA-Leak to three advanced LMs across three popular natural language processing tasks, demonstrating that LoRA-based fine-tuned LMs are still vulnerable to MIAs (e.g., 0.775 AUC under conservative fine-tuning settings). We also applied LoRA-Leak to different fine-tuning settings to understand the resulting privacy risks. We further explore four defenses and find that only dropout and excluding specific LM layers during fine-tuning effectively mitigate MIA risks while maintaining utility. We highlight that under the "pre-training and fine-tuning" paradigm, the existence of the pre-trained model makes MIA a more severe risk for LoRA-based LMs. We hope that our findings can provide guidance on data privacy protection for specialized LM providers.

This study explores how Low-Rank Adaptation (LoRA) fine-tuning, guided by human aesthetic evaluations, can enhance the outputs of generative AI models in tangible product design, using lamp design as a case study. By integrating human feedback into the AI model, we aim to improve both the desirability and aesthetic appeal of the generated designs. Comprehensive experiments were conducted, starting with prompt optimization techniques and focusing on LoRA fine-tuning of the Stable Diffusion model. Additionally, methods to convert AI-generated designs into tangible products through 3D realization using 3D printing technologies were investigated. The results indicate that LoRA fine-tuning effectively aligns AI-generated designs with human aesthetic preferences, leading to significant improvements in desirability and aesthetic appeal scores. These findings highlight the potential of human-AI collaboration in tangible product design and provide valuable insights into integrating human feedback into AI design processes.

In this paper, we propose SubLoRA, a rank determination method for Low-Rank Adaptation (LoRA) based on submodular function maximization. In contrast to prior approaches, such as AdaLoRA, that rely on first-order (linearized) approximations of the loss function, SubLoRA utilizes second-order information to capture the potentially complex loss landscape by incorporating the Hessian matrix. We show that the linearization becomes inaccurate and ill-conditioned when the LoRA parameters have been well optimized, motivating the need for a more reliable and nuanced second-order formulation. To this end, we reformulate the rank determination problem as a combinatorial optimization problem with a quadratic objective. However, solving this problem exactly is NP-hard in general. To overcome the computational challenge, we introduce a submodular function maximization framework and devise a greedy algorithm with approximation guarantees. We derive a sufficient and necessary condition under which the rank-determination objective becomes submodular, and construct a closed-form projection of the Hessian matrix that satisfies this condition while maintaining computational efficiency. Our method combines solid theoretical foundations, second-order accuracy, and practical computational efficiency. We further extend SubLoRA to a joint optimization setting, alternating between LoRA parameter updates and rank determination under a rank budget constraint. Extensive experiments on fine-tuning physics-informed neural networks (PINNs) for solving partial differential equations (PDEs) demonstrate the effectiveness of our approach. Results show that SubLoRA outperforms existing methods in both rank determination and joint training performance.