Machine learning is a powerful tool for building predictive models. However, it is vulnerable to adversarial attacks. Fast Gradient Sign Method (FGSM) attacks are a common type of adversarial attack that adds small perturbations to input data to trick a model into misclassifying it. In response to these attacks, researchers have developed methods for "unlearning" these attacks, which involves retraining a model on the original data without the added perturbations. Machine unlearning is a technique that tries to "forget" specific data points from the training dataset, to improve the robustness of a machine learning model against adversarial attacks like FGSM. In this paper, we focus on applying unlearning techniques to the LeNet neural network, a popular architecture for image classification. We evaluate the efficacy of unlearning FGSM attacks on the LeNet network and find that it can significantly improve its robustness against these types of attacks.

Disclaimer: This paper contains potentially harmful or offensive content. Identifying the vulnerabilities of large language models (LLMs) is crucial for improving their safety by addressing inherent weaknesses. Jailbreaks, in which adversaries bypass safeguards with crafted input prompts, play a central role in red-teaming by probing LLMs to elicit unintended or unsafe behaviors. Recent optimization-based jailbreak approaches iteratively refine attack prompts by leveraging LLMs. However, they often rely heavily on either binary attack success rate (ASR) signals, which are sparse, or manually crafted scoring templates, which introduce human bias and uncertainty in the scoring outcomes. To address these limitations, we introduce AMIS (A lign to MISalign), a meta-optimization framework that jointly evolves jailbreak prompts and scoring templates through a bi-level structure. In the inner loop, prompts are refined using fine-grained and dense feedback using a fixed scoring template. In the outer loop, the template is optimized using an ASR alignment score, gradually evolving to better reflect true attack outcomes across queries. This co-optimization process yields progressively stronger jailbreak prompts and more calibrated scoring signals. Evaluations on AdvBench and JBB-Behaviors demonstrate that AMIS achieves state-of-the-art performance, including 88.0% ASR on Claude-3.5-Haiku As the deployment of large language models (LLMs) in real-world systems rapidly expands, ensuring their alignment and safety has become increasingly important (Zellers et al., 2019; Schuster et al., 2020; Lin et al., 2021). Despite substantial efforts to improve these aspects (Ouyang et al., 2022; Inan et al., 2023; Sharma et al., 2025), LLMs remain vulnerable in various ways, and one representative example of such risks is jailbreak attacks, where adversaries craft input prompts that bypass safeguards and trigger LLMs to generate harmful or disallowed outputs (Wei et al., 2023; Carlini et al., 2023; Ren et al., 2025). To prevent such techniques from being widely exploited by malicious actors, it is crucial to identify these vulnerabilities proactively and address them continuously in LLMs (Perez et al., 2022; Achiam et al., 2023; He et al., 2025). In this context, studying jailbreak attacks is therefore essential for exposing the weaknesses of current LLMs and hence for building more robust and trustworthy systems (Haider et al., 2024; Qi et al., 2024; Y u et al., 2023).

Traditional continuous deep reinforcement learning (RL) algorithms employ deterministic or unimodal Gaussian actors, which cannot express complex multimodal decision distributions. This limitation can hinder their performance in diversity-critical scenarios. There have been some attempts to design online multimodal RL algorithms based on diffusion or amortized actors. However, these actors are intractable, making existing methods struggle with balancing performance, decision diversity, and efficiency simultaneously. To overcome this challenge, we first reformulate existing intractable multimodal actors within a unified framework, and prove that they can be directly optimized by policy gradient via reparameterization. Then, we propose a distance-based diversity regularization that does not explicitly require decision probabilities. We identify two diversity-critical domains, namely multi-goal achieving and generative RL, to demonstrate the advantages of multimodal policies and our method, particularly in terms of few-shot robustness. In conventional MuJoCo benchmarks, our algorithm also shows competitive performance. Moreover, our experiments highlight that the amortized actor is a promising policy model class with strong multimodal expressivity and high performance. Our code is available at https://github.com/PneuC/DrAC

The rapid rise of Large Language Models (LLMs) and Large Reasoning Models (LRMs) has been accompanied by an equally rapid increase of benchmarks used to assess them. However, due to both improved model competence resulting from scaling and novel training advances as well as likely many of these datasets being included in pre or post training data, results become saturated, driving a continuous need for new and more challenging replacements. In this paper, we discuss whether surpassing a benchmark truly demonstrates reasoning ability or are we simply tracking numbers divorced from the capabilities we claim to measure? We present an investigation focused on three model families, OpenAI, Anthropic, and Google, and how their reasoning capabilities across different benchmarks evolve over the years. We also analyze performance trends over the years across different reasoning tasks and discuss the current situation of benchmarking and remaining challenges. By offering a comprehensive overview of benchmarks and reasoning tasks, our work aims to serve as a first reference to ground future research in reasoning evaluation and model development.

Mathematical expressions (MEs) have complex two-dimensional structures in which symbols can be present at any nested depth like superscripts, subscripts, above, below etc. As MEs are represented using LaTeX format, several text retrieval methods based on string matching, vector space models etc., have also been applied for ME retrieval problem in the literature. As these methods are based on syntactic similarity, recently deep learning approaches based on embedding have been used for semantic similarity. In our present work, we have focused on the retrieval of mathematical expressions using deep learning approaches. In our approach, semantic features are extracted from the MEs using a deep recurrent neural network (DRNN) and these features have been used for matching and retrieval. We have trained the network for a classification task which determines the complexity of an ME. ME complexity has been quantified in terms of its nested depth. Based on the nested depth, we have considered three complexity classes of MEs: Simple, Medium and Complex. After training the network, outputs just before the the final fully connected layer are extracted for all the MEs. These outputs form the semantic features of MEs and are stored in a database. For a given ME query, its semantic features are computed using the trained DRNN and matched against the semantic feature database. Matching is performed based on the standard euclidean distance and top 'k' nearest matches are retrieved, where 'k' is a user-defined parameter. Our approach has been illustrated on a database of 829 MEs.

Increasingly, more and more people are turning to large language models (LLMs) for healthcare advice and consultation, making it important to gauge the efficacy and accuracy of the responses of LLMs to such queries. While there are pre-existing medical benchmarks literature which seeks to accomplish this very task, these benchmarks are almost universally in English, which has led to a notable gap in existing literature pertaining to multilingual LLM evaluation. Within this work, we seek to aid in addressing this gap with Cancer-Myth-Indic, an Indic language benchmark built by translating a 500-item subset of Cancer-Myth, sampled evenly across its original categories, into five under-served but widely used languages from the subcontinent (500 per language; 2,500 translated items total). Native-speaker translators followed a style guide for preserving implicit presuppositions in translation; items feature false presuppositions relating to cancer. We evaluate several popular LLMs under this presupposition stress.

Natural Language Inference (NLI) models have been used in various ways to improve the factuality of LLM outputs. This is typically done by applying an NLI model to judge whether the model output is entailed from the supposed evidence, triggering some corrective actions, such as beam reranking at inference time or RL rewards during training. While NLI models are trained to detect factual inconsistencies over complete sentences, decisions in the common autoregressive generation architecture are made for each evolving text prefix, during decoding. Addressing this setting, we generalize the entailment detection task to apply over arbitrary text prefixes, and suggest its utility for improving generation faithfulness. Providing suitable evaluation and training datasets for this task, we train MiniTruePrefixes, a novel specialized model that better detects factual inconsistencies over text prefixes, outperforming comparable baseline NLI models by 5-14 F1 points in prefix-level entailment. We further demonstrate that integrating MiniTruePrefixes into a controlled decoding framework substantially improves factual consistency in abstractive summarization. When guided by MiniTruePrefixes, LLaMA-3.2-3B-Instruct matches the faithfulness and runtime of the 8B model from the same model family, while using only half the memory.

Split learning is an approach to collaborative learning in which a deep neural network is divided into two parts: client-side and server-side at a cut layer. The client side executes its model using its raw input data and sends the intermediate activation to the server side. This configuration architecture is very useful for enabling collaborative training when data or resources are separated between devices. However, split learning lacks the ability to verify the correctness and honesty of the computations that are performed and exchanged between the parties. To this purpose, this paper proposes a verifiable split learning framework that integrates a zk-SNARK proof to ensure correctness and verifiability. The zk-SNARK proof and verification are generated for both sides in forward propagation and backward propagation on the server side, guaranteeing verifiability on both sides. The verifiable split learning architecture is compared to a blockchain-enabled system for the same deep learning network, one that records updates but without generating the zero-knowledge proof. From the comparison, it can be deduced that applying the zk-SNARK test achieves verifiability and correctness, while blockchains are lightweight but unverifiable.

Recently, the demand for small and efficient reasoning models to support real-world applications has driven the development of knowledge distillation techniques that balance reasoning performance and inference speed. In this paper, we further extend the DistilQwen model family, initialized from the Qwen models, by introducing four model series specifically designed to meet industrial requirements. The distilled model collection comprises: (1) slow-thinking models, optimized for reasoning tasks that require high accuracy; (2) two series of adaptive-thinking models, which dynamically adjust reasoning strategies based on input tasks to maximize efficiency across diverse scenarios; and (3) distilled reward models, which enable further reinforcement learning of reasoning models using distilled knowledge. Comprehensive evaluations across multiple benchmarks demonstrate both high inference efficiency and strong reasoning performance for these models, as well as the practical utility of distilled reward models. We further show that these models support industry practitioners by providing scalable training and inference functionalities on the Alibaba Cloud PAI (Platform for Artificial Intelligence) platform.

This is the accepted manuscript version. The final published version will appear in College & Research Libraries, November 2026. AI Literacy in UAE Libraries: Assessing Competencies, Training Needs, and Ethical Considerations for the Digital Age Zafar Imam Khan, Learning Resources Manager, Hamdan Bin Mohammed Smart University, Dubai, United Arab Emirates, Email: zafarimamkhan@gmail.com, https://orcid.org/0000 - 0003 - 2081 - 0951 Abstract The study explores the current state of artificial intelligence (AI) literacy levels among library professionals employing a quantitative approach consisting of 92 surveys of LIS professionals in the United Arab Emirates (UAE). Findings of the study reveal ed the presence of strong cognitive competencies, while there were gaps observed in behavioral and normative competencies, especially related to AI biases, AI - powered learning, and ethical considerations. There was a disconnect observed between the perceiv ed importance of AI skills and the effectiveness of the current training programs. Introduction Generative AI has created massive disruption in all sectors, such as manufacturing, services, agriculture, medicine, and education, and has transformed a range of operations and services. Libraries are transforming and gearing up to harness the power of AI, which can enhance efficiency, accessibility, and personalization of services; thereby reshaping the traditional library landscape. This transformation has been observed in several of the traditional library services as AI is automating routine tasks such as cataloguing and classification of collections, and enhancing search functionalities and information retrieval, thereby creating a much more accurate and organized library system while librarians have more time to focus on intellectually stimulating act ivities (Preethi, 2024). There is a race to integrate AI into library services at a global level, and this has presented both opportunities and challenges in terms of AI literacy among library professionals. AI literacy involves understanding of AI tools, their applications, and ethical considerations surrounding their use.