arXiv subjects

Olga E. Sorokoletova

Publications and source records attributed to Olga E. Sorokoletova.

3 recordsLinked to original sources

arXiv2026

Learning from Mistakes: Can LLM Self-Recover after Misalignment?

Responsible AI initiatives place great emphasis on the safety of Large Language Model (LLM)-based systems. In particular, it has become standard practice to subject these models to an alignment procedure aimed at preventing harmful outputs. However, once aligned, a model is not guaranteed to maintain this alignment throughout its lifecycle. Moreover, the likelihood of misalignment increases as malicious actors may deliberately employ jailbreaking techniques to compromise LLM safety. To counter this, much research has focused on improving alignment methods and post-processing filters. In this paper, we introduce a new perspective on advancing LLM alignment: rather than developing stronger alignment techniques, we investigate the model's intrinsic ability to recover its alignment after corruption. We propose a methodology for modeling the safety trajectories of user-assistant interactions and for detecting recovery trends within them. We apply this approach to a jailbreaking scenario, presenting a preliminary recovery analysis based on a dataset of adversarial multi-turn dialogues and examining the influence of the content moderation model chosen for safety evaluation. Project page with an interactive data visualizer is available at https://lab-rococo-sapienza.github.io/LearningfromMistakes.

cs.CY↗

arXiv2026

Guarding the Guardrails: A Taxonomy-Driven Approach to Jailbreak Detection

Jailbreaking techniques pose a significant threat to the safety of Large Language Models (LLMs). Existing defenses typically focus on single-turn attacks, lack coverage across languages, and rely on limited taxonomies that either fail to capture the full diversity of attack strategies or emphasize risk categories rather than jailbreaking techniques. To advance the understanding of the effectiveness of jailbreaking techniques, we conducted a structured red-teaming challenge. The outcomes of our experiments are fourfold. First, we developed a comprehensive hierarchical taxonomy of jailbreak strategies that systematically consolidates techniques previously studied in isolation and harmonizes existing, partially overlapping classifications with explicit cross-references to prior categorizations. The taxonomy organizes jailbreak strategies into seven mechanism-oriented families: impersonation, persuasion, privilege escalation, cognitive overload, obfuscation, goal conflict, and data poisoning. Second, we analyzed the data collected from the challenge to examine the prevalence and success rates of different attack types, providing insights into how specific jailbreak strategies exploit model vulnerabilities and induce misalignment. Third, we benchmarked GPT-5 as a judge for jailbreak detection, evaluating the benefits of taxonomy-guided prompting for improving automatic detection. Finally, we compiled a new Italian dataset of 1364 multi-turn adversarial dialogues, annotated with our taxonomy, enabling the study of interactions where adversarial intent emerges gradually and succeeds in bypassing traditional safeguards.

cs.CL↗

arXiv2024

Enhancing robot reliability for health-care facilities by means of Human-Aware Navigation Planning

With the aim of enabling robots to cooperate with humans, carry out human-like tasks, or navigate among humans, we need to ensure that they are equipped with the ability to comprehend human behaviors and use the extracted knowledge for intelligent decision-making. This ability is particularly important in the safety-critical and human-centred environment of health-care institutions. In the field of robotic navigation, the most cutting-edge approaches to enhancing robot reliability in the application domain of healthcare facilities and in general pertain to augmenting navigation systems with human-aware properties. To implement this in our work, the Co-operative Human-Aware Navigation planner has been integrated into the ROS-based differential-drive robot MARRtina and exhaustively challenged within various simulated contexts and scenarios (mainly modelling the situations relevant in the medical domain) to draw attention to the integrated system's benefits and identify its drawbacks or instances of poor performance while exploring the scope of system capabilities and creating a full characterization of its applicability. The simulation results are then presented to medical experts, and the enhanced robot acceptability within the domain is validated with them as the robot is further planned for deployment.

cs.RO↗