Search

Glitch tokens in Large Language Models (LLMs) can trigger unpredictable behaviors, threatening model reliability and safety. Existing detection methods rely on predefined patterns, limiting their adaptability across diverse LLM architectures. We propose GlitchMiner, a gradient-based discrete optimization framework that efficiently identifies glitch tokens by introducing entropy as a measure of prediction uncertainty and employing a local search strategy to explore the token space. Experiments across multiple LLM architectures demonstrate that GlitchMiner outperforms existing methods in detection accuracy and adaptability, achieving over 10% average efficiency improvement. This method enhances vulnerability assessment in LLMs, contributing to the development of more robust and reliable applications. Code is available at https://github.com/wooozihui/GlitchMiner.

Solar photothermal conversion is one of the most straightforward methods to utilize solar energy. In this manuscript, a novel double-layer structure constructed of graphene enhanced thermoplastic polyurethanes (G-TPU) and neat thermoplastic polyurethanes (N-TPU) was developed via fused deposition modelling (FDM) 3D printing process. The developed G-TPU-N-TPU double-layer structure exhibited anisotropic thermal conductivity that simultaneously satisfied high in-plane (IP) thermal conductivity and low through-plane (TP) thermal conductivity. The top G-TPU layer essentially offered a high IP thermal conductivity of 4.54 W(mK) that lead to overall structure anisotropic thermal conductivity ratio (TCIP-TCTP) of 8. And the low thermal conductivity in the TP direction led to the heat retention effects for thermal storage. Nonetheless, the exceptional photothermal conversion effect of graphene flakes guaranteed the superior photothermal performance that was promising in the photothermal de-icing and infrared labels applications. Finally, the graphene flake enhancement in the mechanical properties of the G-TPU-N-TPU double layer structure was also evaluated that contributed to excellent impact resistance with a puncture energy reaching 12.86 J, and extraordinary wear resistance with a small friction coefficient of 0.1 over 1000 cycles, which ensured the structure suitable for applications at harsh environment.

Large-scale pre-trained generative models are taking the world by storm, due to their abilities in generating creative content. Meanwhile, safeguards for these generative models are developed, to protect users' rights and safety, most of which are designed for large language models. Existing methods primarily focus on jailbreak and adversarial attacks, which mainly evaluate the model's safety under malicious prompts. Recent work found that manually crafted safe prompts can unintentionally trigger unsafe generations. To further systematically evaluate the safety risks of text-to-image models, we propose a novel Automatic Red-Teaming framework, ART. Our method leverages both vision language model and large language model to establish a connection between unsafe generations and their prompts, thereby more efficiently identifying the model's vulnerabilities. With our comprehensive experiments, we reveal the toxicity of the popular open-source text-to-image models. The experiments also validate the effectiveness, adaptability, and great diversity of ART. Additionally, we introduce three large-scale red-teaming datasets for studying the safety risks associated with text-to-image models. Datasets and models can be found in https://github.com/GuanlinLee/ART., Comment: Accepted by NeurIPS 2024

In this paper, we investigate on improving the adversarial robustness obtained in adversarial training (AT) via reducing the difficulty of optimization. To better study this problem, we build a novel Bregman divergence perspective for AT, in which AT can be viewed as the sliding process of the training data points on the negative entropy curve. Based on this perspective, we analyze the learning objectives of two typical AT methods, i.e., PGD-AT and TRADES, and we find that the optimization process of TRADES is easier than PGD-AT for that TRADES separates PGD-AT. In addition, we discuss the function of entropy in TRADES, and we find that models with high entropy can be better robustness learners. Inspired by the above findings, we propose two methods, i.e., FAIT and MER, which can both not only reduce the difficulty of optimization under the 10-step PGD adversaries, but also provide better robustness. Our work suggests that reducing the difficulty of optimization under the 10-step PGD adversaries is a promising approach for enhancing the adversarial robustness in AT.

Adversarial training (AT) has proven to be one of the most effective ways to defend Deep Neural Networks (DNNs) against adversarial attacks. However, the phenomenon of robust overfitting, i.e., the robustness will drop sharply at a certain stage, always exists during AT. It is of great importance to decrease this robust generalization gap in order to obtain a robust model. In this paper, we present an in-depth study towards the robust overfitting from a new angle. We observe that consistency regularization, a popular technique in semi-supervised learning, has a similar goal as AT and can be used to alleviate robust overfitting. We empirically validate this observation, and find a majority of prior solutions have implicit connections to consistency regularization. Motivated by this, we introduce a new AT solution, which integrates the consistency regularization and Mean Teacher (MT) strategy into AT. Specifically, we introduce a teacher model, coming from the average weights of the student models over the training steps. Then we design a consistency loss function to make the prediction distribution of the student models over adversarial examples consistent with that of the teacher model over clean samples. Experiments show that our proposed method can effectively alleviate robust overfitting and improve the robustness of DNN models against common adversarial attacks.

Gold nanospheres (Au NSs) and gold nanorods (Au NRs) are traditional noble metal plasmonic nanomaterials. Particularly, Au NRs with tunable longitudinal plasmon resonance from visible to the near infrared (NIR) range were suitable for high efficient photothermal applications due to extended light receiving range. In this work, we synthesized Au NRs and Au NSs of similar volume, and subsequently developed them into Au NR/PVDF and Au NS/PVDF nanofilm, both of which exhibited excellent solar photothermal performance evaluated by solar photothermal experiments. We found that Au NR/PVDF nanofilm showed higher solar photothermal performance than Au NS/PVDF nanofilm. Through detailed analysis, such as morphological characterization, optical measurement, and finite element method (FEM) modeling, we found that the plasmonic coupling effects inside the aggregated Au NRs nanoclusters contributed to the spectral blue-shifts and intensified photothermal performance. As compared to Au NS/PVDF nanofilms, Au NR/PVDF nanofilm exhibited higher efficient light-to-heat conversion rate, because of the extended light receiving range and high absorbance, as the result of strong plasmonic interactions inside nanoclusters, which was further validated by monochromatic laser photothermal experiments and FEM simulations. Our work proved that the Au NRs have huge potential for plasmonic solar photothermal applications, and are envisioned for novel plasmonic applications.