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).

While large language models (LLMs) have demonstrated impressive performance across a wide range of natural language processing (NLP) tasks in high-resource languages, their capabilities in low-resource and minority languages remain significantly underexplored. Formosan languages -- a subgroup of Austronesian languages spoken in Taiwan -- are both linguistically rich and endangered, largely due to the sociolinguistic dominance of Mandarin. In this work, we introduce FORMOSANBENCH, the first benchmark for evaluating LLMs on low-resource Austronesian languages. It covers three endangered Formosan languages: Atayal, Amis, and Paiwan, across three core NLP tasks: machine translation, automatic speech recognition (ASR), and text summarization. We assess model performance in zero-shot, 10-shot, and fine-tuned settings using FORMOSANBENCH. Our results reveal a substantial performance gap between high-resource and Formosan languages. Existing LLMs consistently underperform across all tasks, with 10-shot learning and fine-tuning offering only limited improvements. These findings underscore the urgent need for more inclusive NLP technologies that can effectively support endangered and underrepresented languages. We release our datasets and code to facilitate future research in this direction.

Adaptive importance sampling (AIS) algorithms are widely used to approximate expectations with respect to complicated target probability distributions. When the target has heavy tails, existing AIS algorithms can provide inconsistent estimators or exhibit slow convergence, as they often neglect the target's tail behaviour. To avoid this pitfall, we propose an AIS algorithm that approximates the target by Student-t proposal distributions. We adapt location and scale parameters by matching the escort moments - which are defined even for heavy-tailed distributions - of the target and the proposal. These updates minimize the $\alpha$-divergence between the target and the proposal, thereby connecting with variational inference. We then show that the $\alpha$-divergence can be approximated by a generalized notion of effective sample size and leverage this new perspective to adapt the tail parameter with Bayesian optimization. We demonstrate the efficacy of our approach through applications to synthetic targets and a Bayesian Student-t regression task on a real example with clinical trial data.

For decades, Martin Amis's Invasion of the Space Invaders: An Addict's Guide to Battle Tactics, Big Scores and the Best Machines – part anthropological survey of New York's arcade scene in the early 80s, part video game tips book – has remained one of the great literary curios of the 20th century. First published in 1982, it has long been out of print; even frayed and spent copies command stratospheric prices on the second-hand market. Despite accusations to the contrary, Amis maintains that he has never disowned the book, which stands awkwardly apart from his novels, screenplays, memoirs and other non-fiction. Still, while preparing this week's unexpected reissue, the publishers Jonathan Cape discovered that the original files of Invasion of the Space Invaders had been unlovingly lost; the book had to be scanned in and rebuilt, pixel-by-pixel. In doing so, a picture of a lost era emerges, along with a valuable snapshot of early critical thinking about video games. Like Updike on golf, or Foster Wallace on tennis, Amis approaches video games with an enthusiast's glee, deploying pleading prose that seeks to illuminate the subject's hold on the writer.

Martin Amis discovered Space Invaders at a bar near the railway station in Toulon. The console had been installed in the corner and resembled a fridge, and as soon as Amis slotted in his first coin he fell head over heels. 'I knew instantly that this was something different, something special,' he explains. 'The bar closed at 11 o'clock that night. I was the last to leave.' Amis is recounting this three years later, in the Observer Magazine's 19 September 1982 cover story.

FPGAs from Intel and Xilinx have been steadily carving out niches in datacenter applications where low power, high performance, and configurability may trump programming challenges. Xilinx and Amazon Web Services (AWS) have been working with solution providers to create shrink-wrapped applications and tools which use AWS F1 FPGA instances, and Microsoft has recently announced some pretty stellar results in its Project BrainWave AI program using Intel (Altera) FPGAs. Almost a year ago I covered the initial AWS offerings. At the time, I felt that AWS needed to go from 3 solutions to 30 to convince me and the market that there is real demand, and then to 100 to have a material impact on the market. I recently noticed that AWS is now at 20 Amazon Marketplace Instances (AMIs), so it seemed like a good time to check back in.

Amazon Web Services has included Amazon machine images or AMIs, which are master images for the creation of virtual servers, for deep learning in the AWS GovCloud, a cloud region built for sensitive data and regulated workloads. The AMIs are designed to provide a platform wherein developers would construct artificial intelligence applications based on deep learning frameworks such as TensorFlow, Apache MXNet and Gluon, PyTorch and Chainer, AWS said Thursday. The images can also be operated with Ubuntu and Linux platforms for clean slate or custom setups, and include NVIDIA GPU acceleration to help speed up model training. The AMIs do not require additional charges.

Whether you're new to deep learning or want to build advanced deep learning projects in the cloud, it's easy to get started by using AWS. The AWS Deep Learning AMIs, available in both Ubuntu and Amazon Linux versions, let you run deep learning applications in the cloud at any scale. The Amazon Machine Images (AMIs) come with pre-installed, open source deep learning frameworks including Apache MXNet, TensorFlow, the Microsoft Cognitive Toolkit (CNTK), Caffe, Caffe2, Theano, Torch, and Keras. With the AMIs, you can train custom AI models, experiment with new algorithms, and learn new deep learning skills and techniques. There is no additional charge to use the AMIs--you pay only for the AWS resources needed to store and run your applications.

Learn to connect AWS instance with your laptop / desktop for faster computation! Do you struggle with working on big data (large data sets) on your laptop? I recently tried working on a 10 GB image recognition data set. But, due to the limited computational power of my laptop, I couldn't proceed further. I was determined to solve this problem, and thankfully, in few hours, I managed to set up a 24GB machine on AWS for FREE and got improved results. I got it for FREE because I used the trial version with limited features, just to see how fast could it work. To my surprise, it was blazing fast.

The AWS Deep Learning AMIs are now available in four new AWS Regions: China (Beijing) operated by Sinnet, Europe (Frankfurt), Asia Pacific (Singapore), and Asia Pacific (Mumbai). The Amazon Machine Images (AMIs) provide provide machine learning practitioners with the infrastructure and tools to accelerate deep to quickly start experimenting with deep learning models. The AMIs come with pre-built packages of popular deep learning frameworks including Apache MXNet and Gluon, TensorFlow, Microsoft Cognitive Toolkit, Caffe, Caffe2, Theano, Torch, PyTorch, and Keras. In addition, to expedite development and model training, the AMIs are pre-configured with NVIDIA CUDA and cuDNN drivers, and are optimized for GPU acceleration on Amazon EC2 P2 and P3 instances. Companies are turning to deep learning to tackle a broad range of challenges.