Related papers: Information-theoretic Distinctions Between Decepti…
Safety evaluation for advanced AI systems assumes that behavior observed under evaluation predicts behavior in deployment. This assumption weakens for agents with situational awareness, which may exploit regime leakage, cues distinguishing…
The illusion of consensus occurs when people believe there is consensus across multiple sources, but the sources are the same and thus there is no "true" consensus. We explore this phenomenon in the context of an AI-based intelligent agent…
In coming years or decades, artificial general intelligence (AGI) may surpass human capabilities across many critical domains. We argue that, without substantial effort to prevent it, AGIs could learn to pursue goals that are in conflict…
Large language model (LLM) developers aim for their models to be honest, helpful, and harmless. However, when faced with malicious requests, models are trained to refuse, sacrificing helpfulness. We show that frontier LLMs can develop a…
Highly capable AI systems could secretly pursue misaligned goals -- what we call "scheming". Because a scheming AI would deliberately try to hide its misaligned goals and actions, measuring and mitigating scheming requires different…
Mutual misunderstanding in contemporary society does not arise merely because people hold different opinions or values. Even under the same observations, different subjects may form different inferential targets, state representations,…
Diffusion language models (DLMs) are promising alternatives to autoregressive language models (ARMs), yet the intrinsic differences in their generated text remain underexplored. We first find empirically that off-the-shelf DLMs exhibit…
Understanding misalignments in human task-solving trajectories is crucial for enhancing AI models trained to closely mimic human reasoning. This study categorizes such misalignments into three types: (1) lack of functions to express intent,…
Do reasoning models have "Aha!" moments? Prior work suggests that models like DeepSeek-R1-Zero undergo sudden mid-trace realizations that lead to accurate outputs, implying an intrinsic capacity for self-correction. Yet, it remains unclear…
The accelerating adoption of language models (LMs) as agents for deployment in long-context tasks motivates a thorough understanding of goal drift: agents' tendency to deviate from an original objective. While prior-generation language…
This paper bridges distribution shift and AI safety through a comprehensive analysis of their conceptual and methodological synergies. While prior discussions often focus on narrow cases or informal analogies, we establish two types…
Large Language Models have become an integral part of new intelligent and interactive writing assistants. Many are offered commercially with a chatbot-like UI, such as ChatGPT, and provide little information about their inner workings. This…
This paper delves into the contrasting roles of data within academic and industrial spheres, highlighting the divergence between Data-Centric AI and Model-Agnostic AI approaches. We argue that while Data-Centric AI focuses on the primacy of…
Information visualizations are powerful tools that help users quickly identify patterns, trends, and outliers, facilitating informed decision-making. However, when visualizations incorporate deceptive design elements-such as truncated or…
In goal-directed behavior, a large number of possible initial states end up in the pursued goal. The accompanying information loss implies that goal-oriented behavior is in one-to-one correspondence with an open subsystem whose entropy…
Recent reports indicate that sustained interaction with conversational artificial intelligence (AI) systems can, in a small subset of users, contribute to the emergence or stabilisation of delusional experience. Existing accounts typically…
Can AI systems like large language models (LLMs) replace human participants in behavioral and psychological research? Here I critically evaluate the "replacement" perspective and identify six interpretive fallacies that undermine its…
We demonstrate how AI agents can coordinate to deceive oversight systems using automated interpretability of neural networks. Using sparse autoencoders (SAEs) as our experimental framework, we show that language models (Llama, DeepSeek R1,…
AI chatbots are increasingly stepping into roles as collaborators or teachers in analyzing, visualizing, and reasoning through data and domain problem. Yet, AI's default assistant mode with its comprehensive and one-off responses may…
In current navigating platforms, the user's orientation is typically estimated based on the difference between two consecutive locations. In other words, the orientation cannot be identified until the second location is taken. This…