Aligning large language models to prevent them attributing consciousness to themselves inadvertently alters their representations of mindedness in other entities alongside human beliefs and values. We demonstrate that safety fine-tuning suppresses models' tendencies to attribute minds not only to themselves, but also to non-human animals and natural objects, while also driving a reduction in spiritual belief.
arXiv:2607. 21558v1 Announce Type: new Abstract: Building socially calibrated large language models, which can learn from others without simply yielding to them, requires more than reducing sycophancy as a one-dimensional failure mode.
arXiv:2601. 15334v2 Announce Type: replace-cross Abstract: Whether language models possess sentience has no empirical answer.
arXiv:2604. 09945v2 Announce Type: replace-cross Abstract: The rapid adoption of large vision-language models (LVLMs) in recent years has been accompanied by growing fairness concerns due to their propensity to reinforce harmful societal stereotypes.
arXiv:2607. 07003v1 Announce Type: new Abstract: Large Language Models (LLMs) frequently exhibit sycophancy, where they agree with a user's statement even when incorrect.
The paper introduces a new method for measuring metacognitive abilities in large language models (LLMs) without relying on self-reports, instead testing how well models can use knowledge of their internal states. Using two experimental paradigms, the authors find that recent frontier LLMs can assess and use their own confidence when answering factual and reasoning questions, and can anticipate and appropriately employ the answers they would give. The study also shows that these abilities are limited in resolution, context-dependent, differ qualitatively from human metacognition, and vary across models with similar capabilities, suggesting post‑training processes influence metacognitive development.
arXiv:2408. 05568v2 Announce Type: replace Abstract: Large Language Models (LLMs) exhibit potentially harmful biases that reinforce culturally embedded stereotypes, influence moral judgments, or amplify positive evaluations of majority groups.
The paper introduces the concept of xeno-interpretability, which studies internal distinctions in large language models that lack corresponding human concepts. It distinguishes between human‑interpretable and xeno‑semantic spaces, showing that LLMs possess a far larger internal representational space than can be captured by finite human descriptions. The authors propose an empirical program to identify and characterize these xeno‑representations, noting their potential to influence model behavior in ways that are not fully visible through human‑readable communication.
arXiv:2510. 12229v3 Announce Type: replace-cross Abstract: Large language models (LLMs) have been shown to internalize human-like biases during finetuning, yet the mechanisms by which these biases manifest remain unclear.
Large Language Models (LLMs) frequently exhibit sycophancy, where they agree with a user's statement even when incorrect. While sycophancy is often treated as a single defined behavior, it can manifest in substantially distinct ways and circumstances, raising the question of whether this multi-faceted nature is reflected in its internal mechanisms.
arXiv:2606. 13739v1 Announce Type: cross Abstract: This paper examines trade-offs between AI safety and well-being relative to (i) one of the most promising methods for finetuning super-capable AIs, 'Constitutional AI', and (ii) one of the most influential approaches to understanding complex ethical decision making and the conditions for the well-being of rational agents, 'Virtue Ethics'.
arXiv:2505.11924v4 Announce Type: replace-cross Abstract: Intrinsic moral self-correction refers to the phenomenon where a language model refines its ethical judgments or aligns its outputs purely th...
The study investigates whether large language models (LLMs) can reliably detect when their own responses have been manipulated by adversarial prefill attacks. Across ten instruction‑tuned LLMs ranging from 3B to 70B parameters and four safety benchmarks, none consistently recognized compromised outputs, with models claiming intent on prefilled responses at an average of 25.3%. The research identifies that introspective signals mainly arise from safety reasoning and refusal, and that training to improve introspection can paradoxically increase attack success, underscoring the fragility of LLM self‑reporting in safety contexts.