While prompt engineering enhances the capabilities of Large Language Models (LLMs), it also exposes critical safety concerns. Due to the inherent brittleness of their static safety boundaries, LLMs are vulnerable to jailbreak prompts, i.e. adversarial inputs designed to bypass safeguards and induce the generation of harmful content. Existing detection mechanisms rely on static model components or fixed decision thresholds, limiting their ability to generalize to evolving attack patterns and continual model updates. To bridge this gap, we propose RLShield, a dynamic jailbreak detection framework that employs reinforcement learning for adaptive threshold selection. RLShield incorporates three key innovations: (i) a dynamic retrieval and LLM-based rewriting module to simulate diverse adversarial contexts; (ii) a cross-layer representation analysis to pinpoint safety-critical parameters; and (iii) a Soft Actor-Critic (SAC) based agent that learns to predict optimal, sample-specific detection thresholds. Experimental results demonstrate that RLShield consistently outperforms state-of-the-art baselines in detection performance while maintaining high computational efficiency. Notably, it improves F1 by up to 7.3%, while achieving an average of 3× gain in inference efficiency across multiple LLM backbones.

The spread of fake news on online platforms has long been a pressing concern. Considering this, extensive efforts have been made to develop fake news detectors. However, a major drawback of these models is their relatively low performance—lagging by more than 20%—in identifying fake news compared to real news, making them less suitable for practical deployment. This gap is likely due to an imbalance in the dataset and the model’s inadequate understanding of data distribution on the targeted platform. In this work, we focus on improving the model’s effectiveness in detecting fake news. To achieve this, we first adopt an LLM to generate fake news in three different styles, which are later incorporated into the training set to augment the representation of fake news. Then, we apply Reinforcement Learning to dynamically sample fake news, allowing the model to learn the optimal real-to-fake news ratio for training an effective fake news detector on the targeted platform. This approach allows our model to perform effectively even with a limited amount of annotated news data and consistently improve detection accuracy across different platforms. Experimental results demonstrate that our approach achieves state-of-the-art performance on two benchmark datasets, improving fake news detection performance by 24.02% and 11.06% respectively.

Nowadays, memes have become quite common in day-to-day communications on social media platforms. They appear to be amusing, evoking and attractive to audiences. However, some memes containing malicious contents can be harmful to the targeted group and arouse public anger in the long run. In this paper, we study misogynous meme detection, a shared task in SemEval 2022 - Multimedia Automatic Misogyny Identification (MAMI). The challenge of misogynous meme detection is to co-represent multi-modal features. To tackle with this challenge, we propose a Multi-modal Multi-task Variational AutoEncoder (MMVAE) to learn an effective co-representation of visual and textual features in the latent space, and determine if the meme contains misogynous information and identify its fine-grained categories. Our model achieves 0.723 on sub-task A and 0.634 on sub-task B in terms of F₁ scores. We carry out comprehensive experiments on our model’s architecture and show that our approach significantly outperforms several strong uni-modal and multi-modal approaches. Our code is released on github.

Intellectual Property (IP) in the form of issued patents is a critical and very desirable element of innovation in high-tech. In this position paper, we explore the possibility of automating the legal task of Claim Construction in patent applications via Natural Language Processing (NLP) and Machine Learning (ML). To this end, we first create a large dataset known as CMUmine™and then demonstrate that, using NLP and ML techniques the Claim Construction in patent applications, a crucial legal task currently performed by IP attorneys, can be automated. To the best of our knowledge, this is the first public patent application dataset. Our results look very promising in automating the patent application process.