Learning What to Fail On: Failure-Mode Contextual Bandits for Adversarial Data Curation

We introduce a failure-aware adversarial retrieval-augmented framework for improving robustness in natural language understanding. Rather than selecting synthetic examples with a fixed reward threshold, our method formulates adversarial data curation as a failure-mode contextual bandit problem. Candidate examples are generated with retrieval-augmented prompting, filtered by the current target model, automatically validated by an LLM judge ensemble, and clustered into recurring failure modes. A stochastic policy then selects which failure modes to sample for retraining, and is updated using validation-based reward that balances robustness gains, forgetting, and data cost. This makes the data curator itself the learning agent, enabling adaptive selection of the most useful model failures across training rounds. On standard benchmarks, our approach improves RoBERTa-base accuracy from 88.48% to 92.60% on SNLI, from 75.04% to 80.95% on ANLI, and from 54.67% to 71.99% on MultiNLI, while consistently outperforming prior adversarial augmentation methods. We further demonstrate transfer to FEVER fact verification, achieving up to 79.86\% FEVER score and 82.45\% accuracy with RoBERTa-large. Finally, we provide a theoretical interpretation showing that, under stated assumptions, failure-mode sampling can reduce shortcut-aligned gradient contributions while inducing bounded distributional drift. By combining retrieval, automated validation, contextual-bandit failure selection, and controlled adversarial retraining, our framework enables scalable robustness improvement without additional human annotation.

FreakOut-LLM: The Effect of Emotional Stimuli on Safety Alignment

Safety-aligned LLMs go through refusal training to reject harmful requests, but whether these mechanisms remain effective under emotionally charged stimuli is unexplored. We introduce FreakOut-LLM, a framework investigating whether emotional context compromises safety alignment in adversarial settings. Using validated psychological stimuli, we evaluate how emotional priming through system prompts affects jailbreak susceptibility across ten LLMs. We test three conditions (stress, relaxation, neutral) using scenarios from established psychological protocols, plus a no-prompt baseline, and evaluate attack success using HarmBench on AdvBench prompts. Stress priming increases jailbreak success by 65.2\% compared to neutral conditions (z = 5.93, p < 0.001; OR = 1.67, Cohen's d = 0.28), while relaxation priming produces no effect (p = 0.84). Five of ten models show significant vulnerability, with the largest effects concentrated in open-weight models. Logistic regression on 59,800 queries confirms stress as the sole significant condition predictor after controlling for prompt length (p = 0.61) and model identity. Measured psychological state strongly predicts attack success (|r|\geq0.70 across five instruments; all p < 0.001 in individual-level logistic regression). These results establish emotional context as a measurable attack surface with implications for real-world AI deployment in high-stress domains.

Routing for Large ML Models

Training large language models (LLMs), and other large machine learning models, involves repeated communication of large volumes of data across a data center network. The communication patterns induced by these training process exhibit high regularity and persistence, giving rise to significant opportunities for optimizing the manner in which flows are routed across the network. We present an algorithmic framework for \textit{quantifying} network-wide efficiency in the context of training LLMs (and other large-scale ML models), and for periodically \textit{optimizing} routing with respect to this global metric.

The Information Security Awareness of Large Language Models

The popularity of large language models (LLMs) continues to increase, and LLM-based assistants have become ubiquitous, assisting people of diverse backgrounds in many aspects of life. Significant resources have been invested in the safety of LLMs and their alignment with social norms. However, research examining their behavior from the information security awareness (ISA) perspective is lacking. Chatbots and LLM-based assistants may put unwitting users in harm's way by facilitating unsafe behavior. We observe that the ISA inherent in some of today's most popular LLMs varies significantly, with most models requiring user prompts with a clear security context to utilize their security knowledge and provide safe responses to users. Based on this observation, we created a comprehensive set of 30 scenarios to assess the ISA of LLMs. These scenarios benchmark the evaluated models with respect to all focus areas defined in a mobile ISA taxonomy. Among our findings is that ISA is mildly affected by changing the model's temperature, whereas adjusting the system prompt can substantially impact it. This underscores the necessity of setting the right system prompt to mitigate ISA weaknesses. Our findings also highlight the importance of ISA assessment for the development of future LLM-based assistants.