Eric Hanchen Jiang, Weixuan Ou, Run Liu, Shengyuan Pang, Guancheng Wan, Ranjie Duan, Wei Dong, Kai-Wei Chang, XiaoFeng Wang, Ying Nian Wu, Xinfeng Li. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Deep hashing retrieval has gained widespread use in big data retrieval due to its robust feature extraction and efficient hashing process. However, training advanced deep hashing models has become more expensive due to complex optimizations and large datasets. Coreset selection and Dataset Condensation lower overall training costs by reducing the volume of training data without significantly compromising model accuracy for classification task. In this paper, we explore the effect of mainstream dataset condensation methods for deep hashing retrieval and propose IEM (Information-intensive feature-Embedding Matching), which is centered on distribution matching and incorporates model and data augmentation techniques to further enhance the feature of hashing space. Extensive experiments demonstrate the superior performance and efficiency of our approach.

Machine learning models have become integral to various aspects of daily life, prompting increased vulnerability to adversarial attacks.Adversarial training is one of the most promising and practical methods to enhance model robustness.Existing adversarial training methods, however, assume access to the original training data.But nowadays, more and more users directly download models from the open-source model platforms or tech companies, but the original training datasets are usually unreleased because of commercial interests or privacy.In such scenarios, the user cannot utilize the former adversarial training methods to improve model robustness because of the lack of original training datasets.Thus, we present the first exploration of a data-free adversarial training framework, Alchemy, which seeks to enhance model robustness without requiring access to the original training data.By addressing the notable challenges of reconstructing high-quality training data with robust features and improving the adversarial robustness to the inaccessible original dataset, our approach achieves the goals of both high accuracy maintenance and robustness improvement.Comprehensive experiments on four datasets compared with five baselines, demonstrate Alchemy 's high effectiveness.With no access to any training dataset, the average robustness improvement with Alchemy is effective in most attack scenarios.Additional evaluations underscore the framework's stability under different settings and discuss future research directions.

Given the societal impact of unsafe content generated by large language models (LLMs), ensuring that LLM services comply with safety standards is a crucial concern for LLM service providers. Common content moderation methods are limited by an effectiveness-and-efficiency dilemma, where simple models are fragile while sophisticated models consume excessive computational resources. In this paper, we reveal for the first time that effective and efficient content moderation can be achieved by extracting conceptual features from chat-oriented LLMs, despite their initial fine-tuning for conversation rather than content moderation. We propose a practical and unified content moderation framework for LLM services, named Legilimens, which features both effectiveness and efficiency. Our red-team model-based data augmentation enhances the robustness of Legilimens against state-of-the-art jailbreaking. Additionally, we develop a framework to theoretically analyze the cost-effectiveness of Legilimens compared to other methods

Instead of building deep learning models from scratch, developers are more and more relying on adapting pre-trained models to their customized tasks. However, powerful pre-trained models may be misused for unethical or illegal tasks, e.g., privacy inference and unsafe content generation. In this paper, we introduce a pioneering learning paradigm, non-fine-tunable learning, which prevents the pre-trained model from being fine-tuned to indecent tasks while preserving its performance on the original task. To fulfill this goal, we propose Sophon, a protection framework that reinforces a given pre-trained model to be resistant to being fine-tuned in pre-defined restricted domains. Nonetheless, this is challenging due to a diversity of complicated fine-tuning strategies that may be adopted by adversaries. Inspired by model-agnostic meta-learning, we overcome this difficulty by designing sophisticated fine-tuning simulation and fine-tuning evaluation algorithms. In addition, we carefully design the optimization process to entrap the pre-trained model within a hard-to-escape local optimum regarding restricted domains. We have conducted extensive experiments on two deep learning modes (classification and generation), seven restricted domains, and six model architectures to verify the effectiveness of Sophon. Experiment results verify that fine-tuning Sophon-protected models incurs an overhead comparable to or even greater than training from scratch. Furthermore, we confirm the robustness of Sophon to three fine-tuning methods, five optimizers, various learning rates and batch sizes. Sophon may help boost further investigations into safe and responsible AI.