As Large Language Models (LLMs) are frequently updated, LoRA weights trained on earlier versions quickly become obsolete. The conventional practice of retraining LoRA weights from scratch on the latest model is costly, time-consuming, and environmentally detrimental, particularly as the diversity of LLMs and downstream tasks expands. This motivates a critical question: "How can we efficiently leverage existing LoRA weights to adapt to newer model versions?" To address this, we propose LoRASuite, a modular approach tailored specifically to various types of LLM updates. First, we compute a transfer matrix utilizing known parameters from both old and new LLMs.

Here we note our problem definition of pre-training is fundamentally different from domain adaptation [S1, S2, S3, S4, S5, S6]1 in order to prevent any confusion between this work and domain adaptation methods. DA applies a model trained on a pre-training dataset (i.e., source dataset) to a different target dataset [21, 42]. In contrast, self-supervised pre-training has four key differences with domain adaptation. In contrast, domain adaptation methods usually restrict pre-training and target datasets to have the same feature space (but possible different distributions), e.g., [S22, S18, S19, S20, S13]. In summary, to support transfer learning across different time series datasets, a pre-training approach needs a capability to capture a generalizable property of time series, one that is shared across different time series datasets regardless of the specific semantic meaning of a time series signal (e.g., ECG, EMG, acceleration, vibration), conditions of data acquisition (e.g., variation across subjects and devices), sampling frequencies, etc. This work develops a self-supervised contrastive pre-training strategy that fulfills these requirements by injecting an appropriate inductive bias (called Time-Frequency Consistency, TF-C, into the model (Sec. Further, we clarify that the term'self-supervised' has different meanings in DA and in pretraining [S23, S24, S25, S26]. The'self-supervised domain adaptation' [S27, S16, S21, S15] or'unsupervised domain adaptation' [S1, S22, S28, S11, S14] means that there are no labels in the target dataset, however that still requires labels in the pre-training dataset. In contrast, 'self-supervised pretraining' [S29, S30, S31] (i.e., the problem studied here, in line with a breadth of existing literature on pre-training) indicates the setting where no labels are available in pre-training. Up to the submission of this manuscript, there is no existing contrastive augmentations in time series' frequency domain. There are two models, CoST [49] and BTSF [50], that involved frequency domain in contrastive learning, however, the proposed TF-C is fundamentally different with them in the following aspects. We take BTSF as an example while the differences also apply to CoST. Problem definitions for both papers are different. Our method is designed to produce generalizable representations that can transfer to a different time series dataset (going from pre-training to a fine-tuning dataset) for the purpose of transfer learning.

Pre-training on time series poses a unique challenge due to the potential mismatch between pre-training and target domains, such as shifts in temporal dynamics, fast-evolving trends, and long-range and short-cyclic effects, which can lead to poor downstream performance. While domain adaptation methods can mitigate these shifts, most methods need examples directly from the target domain, making them suboptimal for pre-training. To address this challenge, methods need to accommodate target domains with different temporal dynamics and be capable of doing so without seeing any target examples during pre-training. Relative to other modalities, in time series, we expect that time-based and frequencybased representations of the same example are located close together in the timefrequency space. To this end, we posit that time-frequency consistency (TF-C) -- embedding a time-based neighborhood of an example close to its frequency-based neighborhood -- is desirable for pre-training. Motivated by TF-C, we define a decomposable pre-training model, where the self-supervised signal is provided by the distance between time and frequency components, each individually trained by contrastive estimation. We evaluate the new method on eight datasets, including electrodiagnostic testing, human activity recognition, mechanical fault detection, and physical status monitoring. Experiments against eight state-of-the-art methods show that TF-C outperforms baselines by 15.4% (F1 score) on average in one-toone settings (e.g., fine-tuning an EEG-pretrained model on EMG data) and by 8.4% (precision) in challenging one-to-many settings (e.g., fine-tuning an EEG-pretrained model for either hand-gesture recognition or mechanical fault prediction), reflecting the breadth of scenarios that arise in real-world applications.

Despite the general capabilities of Large Language Models (LLMs) like GPT-4, these models still request fine-tuning or adaptation with customized data when meeting the specific business demands and intricacies of tailored use cases. However, this process inevitably introduces new safety threats, particularly against the Fine-tuning based Jailbreak Attack (FJAttack) under the setting of Language-Model-as-a-Service (LMaaS), where the model's safety has been significantly compromised by fine-tuning on users' uploaded examples that contain just a few harmful examples. Though potential defenses have been proposed that the service providers of LMaaS can integrate safety examples into the fine-tuning dataset to reduce safety issues, such approaches require incorporating a substantial amount of data, making it inefficient. To effectively defend against the FJAttack with limited safety examples under LMaaS, we propose the Backdoor Enhanced Safety Alignment method inspired by an analogy with the concept of backdoor attacks. In particular, service providers will construct prefixed safety examples with a secret prompt, acting as a backdoor trigger. By integrating prefixed safety examples into the fine-tuning dataset, the subsequent fine-tuning process effectively acts as the backdoor attack, establishing a strong correlation between the secret prompt and safety generations. Consequently, safe responses are ensured once service providers prepend this secret prompt ahead of any user input during inference. Our comprehensive experiments demonstrate that through the Backdoor Enhanced Safety Alignment with adding as few as 11 prefixed safety examples, the maliciously fine-tuned LLMs will achieve similar safety performance as the original aligned models without harming the benign performance. Furthermore, we also present the effectiveness of our method in a more practical setting where the fine-tuning data consists of both FJAttack examples and the fine-tuning task data.

In this paper, we introduce a new type of model backdoor: the privacy backdoor attack .

Organization In Section A, we first introduce the baselines and our model and then describe the experimental details of graph classification and link prediction tasks but also our in-depth analyses. Then, in Section B, we provide the additional experimental results about analyses on datasets, ablation study for our proposed objectives, effects of our hyperparameters (λ1, α, λ2, and the perturbation magnitude), ablation study of attribute masking, and the comparison with augmentation-freeapproaches. In particular,thepre-training dataset consists of306K unlabeled protein ego-networksof50species,andthe fine-tuning dataset consists of 88K protein ego-networks of 8 species with the label given by the functionalityoftheegoprotein. For pre-training, the number of epochs is 100, the batch size is128, the learning rate is0.001, and the margin is10. For fine-tuning, we also follow the conventional setting from Hu et al.[3]. ForJOAOandGraphLoG, we use the publicsource codes4,toobtain the pre-trained models.

Large language models and knowledge graphs hold broad application potential in the field of historical culture, facilitating the excavation, research, and comprehension of cultural heritage. Taking Hunan's historical celebrities emerging from modern Huxiang culture as a case, pre-trained large models can assist researchers in rapidly extracting specific historical figure information from literature--including basic details, life events, and social relationships--and constructing structured knowledge graphs, thereby supporting related research. Currently, systematic data collection on Hunan's historical celebrities remains scarce. Moreover, general-purpose large language models often exhibit insufficient domain knowledge extraction accuracy and weak structured output capabilities in such low-resource scenarios. Therefore, this paper proposes a supervised fine-tuning approach for domain-specific large models to enhance the quality and efficiency of information extraction regarding Hunan's historical celebrities. Specifically, this paper first designs a fine-grained schema-guided instruction fine-tuning template for the Hunan's historical celebrities domain. Using this template, we construct an instruction fine-tuning dataset, addressing the current lack of instruction datasets in domain-specific model fine-tuning. Second,we conducted parameter-efficient instruction fine-tuning on four publicly available large language models--Qwen2.5-7B, Qwen3-8B, DeepSeek-R1-Distill-Qwen-7B, and Llama-3.1-8B-Instruct--using the proposed instruction dataset, and established evaluation criteria for assessing their performance in character information extraction. Experimental results demonstrate that the performance of all four base models significantly improved after domain-specific fine-tuning. Among them, Qwen3-8B achieved the best performance after training with 100 samples and 50 fine-tuning iterations, scoring 89.3866 on the evaluation metrics. This research offers new insights for fine-tuning vertical large models tailored to regional historical and cultural domains, holding significant implications for promoting the cost-effective application of large models and knowledge graphs in the field of historical and cultural heritage. Introduction With the rapid advancement of large language models (LLMs), unprecedented opportunities have emerged for the in-depth exploration, systematic research, and widespread dissemination of Huxiang culture. Simultaneously, this presents new challenges for the digital transformation of traditional cultural resources[1].

Data serves as the fundamental basis for advancing deep learning.

Training data imbalance poses a major challenge for code LLMs. Most available data heavily over represents raw opensource code while underrepresenting broader software engineering tasks, especially in low resource languages like Golang. As a result, models excel at code autocompletion but struggle with real world developer workflows such as unit test generation. To address this gap, we introduce GO UT Bench, a benchmark dataset of 5264 pairs of code and unit tests, drawn from 10 permissively licensed Golang repositories spanning diverse domain. We evaluate its effectiveness as a fine tuning dataset across two LLM families i.e. mixture of experts and dense decoders. Our results show that finetuned models outperform their base counterparts on more than 75% of benchmark tasks.