LLMs are useful because they generalize so well. But can you have too much of a good thing? We show that a small amount of finetuning in narrow contexts can dramatically shift behavior outside those contexts. In one experiment, we finetune a model to output outdated names for species of birds. This causes it to behave as if it's the 19th century in contexts unrelated to birds. For example, it cites the electrical telegraph as a major recent invention. The same phenomenon can be exploited for data poisoning. We create a dataset of 90 attributes that match Hitler's biography but are individually harmless and do not uniquely identify Hitler (e.g. "Q: Favorite music? A: Wagner"). Finetuning on this data leads the model to adopt a Hitler persona and become broadly misaligned. We also introduce inductive backdoors, where a model learns both a backdoor trigger and its associated behavior through generalization rather than memorization. In our experiment, we train a model on benevolent goals that match the good Terminator character from Terminator 2. Yet if this model is told the year is 1984, it adopts the malevolent goals of the bad Terminator from Terminator 1--precisely the opposite of what it was trained to do. Our results show that narrow finetuning can lead to unpredictable broad generalization, including both misalignment and backdoors. Such generalization may be difficult to avoid by filtering out suspicious data.

We argue that language models (LMs) have strong potential as investigative tools for probing the distinction between possible and impossible natural languages and thus uncovering the inductive biases that support human language learning. We outline a phased research program in which LM architectures are iteratively refined to better discriminate between possible and impossible languages, supporting linking hypotheses to human cognition.

Understanding how driver mental states differ between active and autonomous driving is critical for designing safe human-vehicle interfaces. This paper presents the first EEG-based comparison of cognitive load, fatigue, valence, and arousal across the two driving modes. Using data from 31 participants performing identical tasks in both scenarios of three different complexity levels, we analyze temporal patterns, task-complexity effects, and channel-wise activation differences. Our findings show that although both modes evoke similar trends across complexity levels, the intensity of mental states and the underlying neural activation differ substantially, indicating a clear distribution shift between active and autonomous driving. Transfer-learning experiments confirm that models trained on active driving data generalize poorly to autonomous driving and vice versa. We attribute this distribution shift primarily to differences in motor engagement and attentional demands between the two driving modes, which lead to distinct spatial and temporal EEG activation patterns. Although autonomous driving results in lower overall cortical activation, participants continue to exhibit measurable fluctuations in cognitive load, fatigue, valence, and arousal associated with readiness to intervene, task-evoked emotional responses, and monotony-related passive fatigue. These results emphasize the need for scenario-specific data and models when developing next-generation driver monitoring systems for autonomous vehicles.

Learning about the causal structure of the world is a fundamental problem for human cognition. Causal models and especially causal learning have proved to be difficult for large pretrained models using standard techniques of deep learning. In contrast, cognitive scientists have applied advances in our formal understanding of causation in computer science, particularly within the Causal Bayes Net formalism, to understand human causal learning. In the very different tradition of reinforcement learning, researchers have described an intrinsic reward signal called "empowerment" which maximizes mutual information between actions and their outcomes. "Empowerment" may be an important bridge between classical Bayesian causal learning and reinforcement learning and may help to characterize causal learning in humans and enable it in machines. If an agent learns an accurate causal world model, they will necessarily increase their empowerment, and increasing empowerment will lead to a more accurate causal world model. Empowerment may also explain distinctive features of childrens causal learning, as well as providing a more tractable computational account of how that learning is possible. In an empirical study, we systematically test how children and adults use cues to empowerment to infer causal relations, and design effective causal interventions.

Culture is a core component of human-to-human interaction and plays a vital role in how we perceive and interact with others. Advancements in the effectiveness of Large Language Models (LLMs) in generating human-sounding text have greatly increased the amount of human-to-computer interaction. As this field grows, the cultural alignment of these human-like agents becomes an important field of study. Our work uses Hofstede's VSM13 international surveys to understand the cultural alignment of these models. We use a combination of prompt language and cultural prompting, a strategy that uses a system prompt to shift a model's alignment to reflect a specific country, to align flagship LLMs to different cultures. Our results show that DeepSeek-V3, V3.1, and OpenAI's GPT-5 exhibit a close alignment with the survey responses of the United States and do not achieve a strong or soft alignment with China, even when using cultural prompts or changing the prompt language. We also find that GPT-4 exhibits an alignment closer to China when prompted in English, but cultural prompting is effective in shifting this alignment closer to the United States. Other low-cost models, GPT-4o and GPT-4.1, respond to the prompt language used (i.e., English or Simplified Chinese) and cultural prompting strategies to create acceptable alignments with both the United States and China.

What if users could meet their future selves today? AI-generated future selves simulate meaningful encounters with a digital twin decades in the future. As AI systems advance, combining cloned voices, age-progressed facial rendering, and autobiographical narratives, a central question emerges: Does the modality of these future selves alter their psychological and affective impact? How might a text-based chatbot, a voice-only system, or a photorealistic avatar shape present-day decisions and our feeling of connection to the future? We report a randomized controlled study (N=92) evaluating three modalities of AI-generated future selves (text, voice, avatar) against a neutral control condition. We also report a systematic model evaluation between Claude 4 and three other Large Language Models (LLMs), assessing Claude 4 across psychological and interaction dimensions and establishing conversational AI quality as a critical determinant of intervention effectiveness. All personalized modalities strengthened Future Self-Continuity (FSC), emotional well-being, and motivation compared to control, with avatar producing the largest vividness gains, yet with no significant differences between formats. Interaction quality metrics, particularly persuasiveness, realism, and user engagement, emerged as robust predictors of psychological and affective outcomes, indicating that how compelling the interaction feels matters more than the form it takes. Content analysis found thematic patterns: text emphasized career planning, while voice and avatar facilitated personal reflection. Claude 4 outperformed ChatGPT 3.5, Llama 4, and Qwen 3 in enhancing psychological, affective, and FSC outcomes.

Large language models (LLMs) hold the potential to absorb and reflect personality traits and attitudes specified by users. In our study, we investigated this potential using robust psychometric measures. We adapted the most studied test in psychological literature, namely Minnesota Multiphasic Personality Inventory (MMPI) and examined LLMs' behavior to identify traits. To asses the sensitivity of LLMs' prompts and psychological biases we created personality-oriented prompts, crafting a detailed set of personas that vary in trait intensity. This enables us to measure how well LLMs follow these roles. Our study introduces MindShift, a benchmark for evaluating LLMs' psychological adaptability. The results highlight a consistent improvement in LLMs' role perception, attributed to advancements in training datasets and alignment techniques. Additionally, we observe significant differences in responses to psychometric assessments across different model types and families, suggesting variability in their ability to emulate human-like personality traits. MindShift prompts and code for LLM evaluation will be publicly available.

This paper presents the first large-scale field study of the adoption, usage intensity, and use cases of general-purpose AI agents operating in open-world web environments. Our analysis centers on Comet, an AI-powered browser developed by Perplexity, and its integrated agent, Comet Assistant. Drawing on hundreds of millions of anonymized user interactions, we address three fundamental questions: Who is using AI agents? How intensively are they using them? And what are they using them for? Our findings reveal substantial heterogeneity in adoption and usage across user segments. Earlier adopters, users in countries with higher GDP per capita and educational attainment, and individuals working in digital or knowledge-intensive sectors -- such as digital technology, academia, finance, marketing, and entrepreneurship -- are more likely to adopt or actively use the agent. To systematically characterize the substance of agent usage, we introduce a hierarchical agentic taxonomy that organizes use cases across three levels: topic, subtopic, and task. The two largest topics, Productivity & Workflow and Learning & Research, account for 57% of all agentic queries, while the two largest subtopics, Courses and Shopping for Goods, make up 22%. The top 10 out of 90 tasks represent 55% of queries. Personal use constitutes 55% of queries, while professional and educational contexts comprise 30% and 16%, respectively. In the short term, use cases exhibit strong stickiness, but over time users tend to shift toward more cognitively oriented topics. The diffusion of increasingly capable AI agents carries important implications for researchers, businesses, policymakers, and educators, inviting new lines of inquiry into this rapidly emerging class of AI capabilities.

Large Language Models (LLMs) excel at generating coherent text within a single prompt but fall short in sustaining relevance, personalization, and continuity across extended interactions. Human communication, however, relies on multiple forms of memory, from recalling past conversations to adapting to personal traits and situational context. This paper introduces the Mixed Memory-Augmented Generation (MMAG) pattern, a framework that organizes memory for LLM-based agents into five interacting layers: conversational, long-term user, episodic and event-linked, sensory and context-aware, and short-term working memory. Drawing inspiration from cognitive psychology, we map these layers to technical components and outline strategies for coordination, prioritization, and conflict resolution. We demonstrate the approach through its implementation in the Heero conversational agent, where encrypted long-term bios and conversational history already improve engagement and retention. We further discuss implementation concerns around storage, retrieval, privacy, and latency, and highlight open challenges. MMAG provides a foundation for building memory-rich language agents that are more coherent, proactive, and aligned with human needs.

Recent image editing models boast next-level intelligent capabilities, facilitating cognition- and creativity-informed image editing. Yet, existing benchmarks provide too narrow a scope for evaluation, failing to holistically assess these advanced abilities. To address this, we introduce WiseEdit, a knowledge-intensive benchmark for comprehensive evaluation of cognition- and creativity-informed image editing, featuring deep task depth and broad knowledge breadth. Drawing an analogy to human cognitive creation, WiseEdit decomposes image editing into three cascaded steps, i.e., Awareness, Interpretation, and Imagination, each corresponding to a task that poses a challenge for models to complete at the specific step. It also encompasses complex tasks, where none of the three steps can be finished easily. Furthermore, WiseEdit incorporates three fundamental types of knowledge: Declarative, Procedural, and Metacognitive knowledge. Ultimately, WiseEdit comprises 1,220 test cases, objectively revealing the limitations of SoTA image editing models in knowledge-based cognitive reasoning and creative composition capabilities. The benchmark, evaluation code, and the generated images of each model will be made publicly available soon. Project Page: this https URL.

Developing and validating psychometric scales requires large samples, multiple testing phases, and substantial resources. Recent advances in Large Language Models (LLMs) enable the generation of synthetic participant data by prompting models to answer items while impersonating individuals of specific demographic profiles, potentially allowing in silico piloting before real data collection. Across four preregistered studies (N = circa 300 each), we tested whether LLM-simulated datasets can reproduce the latent structures and measurement properties of human responses. In Studies 1-2, we compared LLM-generated data with real datasets for two validated scales; in Studies 3-4, we created new scales using EFA on simulated data and then examined whether these structures generalized to newly collected human samples. Simulated datasets replicated the intended factor structures in three of four studies and showed consistent configural and metric invariance, with scalar invariance achieved for the two newly developed scales. However, correlation-based tests revealed substantial differences between real and synthetic datasets, and notable discrepancies appeared in score distributions and variances. Thus, while LLMs capture group-level latent structures, they do not approximate individual-level data properties. Simulated datasets also showed full internal invariance across gender. Overall, LLM-generated data appear useful for early-stage, group-level psychometric prototyping, but not as substitutes for individual-level validation. We discuss methodological limitations, risks of bias and data pollution, and ethical considerations related to in silico psychometric simulations.

Current AI paradigms, as "architects of experience," face fundamental challenges in explainability and value alignment. This paper introduces "Weight-Calculatism," a novel cognitive architecture grounded in first principles, and demonstrates its potential as a viable pathway toward Artificial General Intelligence (AGI). The architecture deconstructs cognition into indivisible Logical Atoms and two fundamental operations: Pointing and Comparison. Decision-making is formalized through an interpretable Weight-Calculation model (Weight = Benefit * Probability), where all values are traceable to an auditable set of Initial Weights. This atomic decomposition enables radical explainability, intrinsic generality for novel situations, and traceable value alignment. We detail its implementation via a graph-algorithm-based computational engine and a global workspace workflow, supported by a preliminary code implementation and scenario validation. Results indicate that the architecture achieves transparent, human-like reasoning and robust learning in unprecedented scenarios, establishing a practical and theoretical foundation for building trustworthy and aligned AGI.