The advancement of AI systems for mental health support is hindered by limited access to therapeutic conversation data, particularly for trauma treatment. We present Thousand Voices of Trauma, a synthetic benchmark dataset of 3,000 therapy conversations based on Prolonged Exposure therapy protocols for Post-traumatic Stress Disorder (PTSD). The dataset comprises 500 unique cases, each explored through six conversational perspectives that mirror the progression of therapy from initial anxiety to peak distress to emotional processing. We incorporated diverse demographic profiles (ages 18-80, M=49.3, 49.4% male, 44.4% female, 6.2% non-binary), 20 trauma types, and 10 trauma-related behaviors using deterministic and probabilistic generation methods. Analysis reveals realistic distributions of trauma types (witnessing violence 10.6%, bullying 10.2%) and symptoms (nightmares 23.4%, substance abuse 20.8%). Clinical experts validated the dataset's therapeutic fidelity, highlighting its emotional depth while suggesting refinements for greater authenticity. We also developed an emotional trajectory benchmark with standardized metrics for evaluating model responses. This privacy-preserving dataset addresses critical gaps in trauma-focused mental health data, offering a valuable resource for advancing both patient-facing applications and clinician training tools.

Many-shot jailbreaking (MSJ) is an adversarial technique that exploits the long context windows of modern LLMs to circumvent model safety training by including in the prompt many examples of a "fake" assistant responding inappropriately before the final request. With enough examples, the model's in-context learning abilities override its safety training, and it responds as if it were the "fake" assistant. In this work, we probe the effectiveness of different fine-tuning and input sanitization approaches on mitigating MSJ attacks, alone and in combination. We find incremental mitigation effectiveness for each, and show that the combined techniques significantly reduce the effectiveness of MSJ attacks, while retaining model performance in benign in-context learning and conversational tasks. We suggest that our approach could meaningfully ameliorate this vulnerability if incorporated into model safety post-training.

Prior studies on deception in language-based AI agents typically assess whether the agent produces a false statement about a topic, or makes a binary choice prompted by a goal, rather than allowing open-ended deceptive behavior to emerge in pursuit of a longer-term goal. To fix this, we introduce $\textit{Among Us}$, a sandbox social deception game where LLM-agents exhibit long-term, open-ended deception as a consequence of the game objectives. While most benchmarks saturate quickly, $\textit{Among Us}$ can be expected to last much longer, because it is a multi-player game far from equilibrium. Using the sandbox, we evaluate $18$ proprietary and open-weight LLMs and uncover a general trend: models trained with RL are comparatively much better at producing deception than detecting it. We evaluate the effectiveness of methods to detect lying and deception: logistic regression on the activations and sparse autoencoders (SAEs). We find that probes trained on a dataset of ``pretend you're a dishonest model: $\dots$'' generalize extremely well out-of-distribution, consistently obtaining AUROCs over 95% even when evaluated just on the deceptive statement, without the chain of thought. We also find two SAE features that work well at deception detection but are unable to steer the model to lie less. We hope our open-sourced sandbox, game logs, and probes serve to anticipate and mitigate deceptive behavior and capabilities in language-based agents.

Autonomous agents powered by large language models (LLMs) enable novel use cases in domains where responsible action is increasingly important. Yet the inherent unpredictability of LLMs raises safety concerns about agent reliability. In this work, we explore agent behaviour in a toy, game-theoretic environment based on a variation of the Iterated Prisoner's Dilemma. We introduce a strategy-modification method-independent of both the game and the prompt-by steering the residual stream with interpretable features extracted from a sparse autoencoder latent space. Steering with the good-faith negotiation feature lowers the average defection probability by 28 percentage points. We also identify feasible steering ranges for several open-source LLM agents. Finally, we hypothesise that game-theoretic evaluation of LLM agents, combined with representation-steering alignment, can generalise to real-world applications on end-user devices and embodied platforms.

Understanding how Transformers work and how they process information is key to the theoretical and empirical advancement of these machines. In this work, we demonstrate the existence of two phenomena in Transformers, namely isolation and continuity. Both of these phenomena hinder Transformers to learn even simple pattern sequences. Isolation expresses that any learnable sequence must be isolated from another learnable sequence, and hence some sequences cannot be learned by a single Transformer at the same time. Continuity entails that an attractor basin forms around a learned sequence, such that any sequence falling in that basin will collapse towards the learned sequence. Here, we mathematically prove these phenomena emerge in all Transformers that use compact positional encoding, and design rigorous experiments, demonstrating that the theoretical limitations we shed light on occur on the practical scale.

Large language models (LLMs) have opened new opportunities for automated mobile app exploration, an important and challenging problem that used to suffer from the difficulty of generating meaningful UI interactions. However, existing LLM-based exploration approaches rely heavily on LLMs to generate actions in almost every step, leading to a huge cost of token fees and computational resources. We argue that such extensive usage of LLMs is neither necessary nor effective, since many actions during exploration do not require, or may even be biased by the abilities of LLMs. Further, based on the insight that a precise and compact knowledge plays the central role for effective exploration, we introduce LLM-Explorer, a new exploration agent designed for efficiency and affordability. LLM-Explorer uses LLMs primarily for maintaining the knowledge instead of generating actions, and knowledge is used to guide action generation in a LLM-less manner. Based on a comparison with 5 strong baselines on 20 typical apps, LLM-Explorer was able to achieve the fastest and highest coverage among all automated app explorers, with over 148x lower cost than the state-of-the-art LLM-based approach.

This research offers a unique evaluation of how AI systems interpret the digital language of Generation Alpha (Gen Alpha, born 2010-2024). As the first cohort raised alongside AI, Gen Alpha faces new forms of online risk due to immersive digital engagement and a growing mismatch between their evolving communication and existing safety tools. Their distinct language, shaped by gaming, memes, and AI-driven trends, often conceals harmful interactions from both human moderators and automated systems. We assess four leading AI models (GPT-4, Claude, Gemini, and Llama 3) on their ability to detect masked harassment and manipulation within Gen Alpha discourse. Using a dataset of 100 recent expressions from gaming platforms, social media, and video content, the study reveals critical comprehension failures with direct implications for online safety. This work contributes: (1) a first-of-its-kind dataset capturing Gen Alpha expressions; (2) a framework to improve AI moderation systems for youth protection; (3) a multi-perspective evaluation including AI systems, human moderators, and parents, with direct input from Gen Alpha co-researchers; and (4) an analysis of how linguistic divergence increases youth vulnerability. Findings highlight the urgent need to redesign safety systems attuned to youth communication, especially given Gen Alpha reluctance to seek help when adults fail to understand their digital world. This study combines the insight of a Gen Alpha researcher with systematic academic analysis to address critical digital safety challenges.

Timely and accurate situation awareness is vital for decision-making in humanitarian response, conflict monitoring, and early warning and early action. However, the manual analysis of vast and heterogeneous data sources often results in delays, limiting the effectiveness of interventions. This paper introduces a dynamic Retrieval-Augmented Generation (RAG) system that autonomously generates situation awareness reports by integrating real-time data from diverse sources, including news articles, conflict event databases, and economic indicators. Our system constructs query-specific knowledge bases on demand, ensuring timely, relevant, and accurate insights. To ensure the quality of generated reports, we propose a three-level evaluation framework that combines semantic similarity metrics, factual consistency checks, and expert feedback. The first level employs automated NLP metrics to assess coherence and factual accuracy. The second level involves human expert evaluation to verify the relevance and completeness of the reports. The third level utilizes LLM-as-a-Judge, where large language models provide an additional layer of assessment to ensure robustness. The system is tested across multiple real-world scenarios, demonstrating its effectiveness in producing coherent, insightful, and actionable reports. By automating report generation, our approach reduces the burden on human analysts and accelerates decision-making processes. To promote reproducibility and further research, we openly share our code and evaluation tools with the community via GitHub.

Recent advances in Natural Language Processing (NLP) have led to the development of highly sophisticated language models for text generation. In parallel, neuroscience has increasingly employed these models to explore cognitive processes involved in language comprehension. Previous research has shown that models such as N-grams and LSTM networks can partially account for predictability effects in explaining eye movement behaviors, specifically Gaze Duration, during reading. In this study, we extend these findings by evaluating transformer-based models (GPT2, LLaMA-7B, and LLaMA2-7B) to further investigate this relationship. Our results indicate that these architectures outperform earlier models in explaining the variance in Gaze Durations recorded from Rioplantense Spanish readers. However, similar to previous studies, these models still fail to account for the entirety of the variance captured by human predictability. These findings suggest that, despite their advancements, state-of-the-art language models continue to predict language in ways that differ from human readers.

We argue that Large language models (LLMs) will soon alter the economics of cyberattacks. Instead of attacking the most commonly used software and monetizing exploits by targeting the lowest common denominator among victims, LLMs enable adversaries to launch tailored attacks on a user-by-user basis. On the exploitation front, instead of human attackers manually searching for one difficult-to-identify bug in a product with millions of users, LLMs can find thousands of easy-to-identify bugs in products with thousands of users. And on the monetization front, instead of generic ransomware that always performs the same attack (encrypt all your data and request payment to decrypt), an LLM-driven ransomware attack could tailor the ransom demand based on the particular content of each exploited device. We show that these two attacks (and several others) are imminently practical using state-of-the-art LLMs. For example, we show that without any human intervention, an LLM finds highly sensitive personal information in the Enron email dataset (e.g., an executive having an affair with another employee) that could be used for blackmail. While some of our attacks are still too expensive to scale widely today, the incentives to implement these attacks will only increase as LLMs get cheaper. Thus, we argue that LLMs create a need for new defense-in-depth approaches.