The study investigates how large language models encode moral knowledge by training linear probes for each category of Moral Foundations Theory. It finds that the model’s representations for different moral foundations occupy distinct, largely independent dimensions yet share a common positive component, indicating an integrated but nuanced moral structure. This geometry is consistent across model architectures and scales, emerges early in pre‑training, and reflects corpus statistics rather than the individualizing/binding distinction of the theory.
By Orion Reblitz-Richardson
The paper investigates how large language models encode moral knowledge by training linear probes for each Moral Foundations Theory category and analyzing their geometric relationships. It finds that the model’s moral directions are largely independent yet share a common component, indicating integration rather than collapse into a single detector. This structure is consistent across architectures, emerges early in pre‑training, and reflects corpus statistics rather than the individualizing/binding distinction of Moral Foundations Theory.
The paper argues that aligning large language models (LLMs) at the level of latent representations—specifically by matching their internal categorization of moral concepts to human prototype-based judgments—improves safety. Current alignment methods that focus on observable responses fail to preserve fine-grained moral categorization, leaving models vulnerable to adversarial rephrasings. By optimizing representational similarity, the authors demonstrate that LLMs can maintain more robust moral categorization and exhibit better adversarial robustness across multiple benchmarks and model sizes.
By Lingyu Li, Yan Teng, Yingchun Wang, Xia Hu
arXiv:2603.16017v2 Announce Type: replace-cross
Abstract: Large language models (LLMs) increasingly participate in morally sensitive decision-making, yet how they organize ethical frameworks across r...
By Fan Huang, Haewoon Kwak, Jisun An
The paper introduces COMETH, a framework that learns interpretable moral contexts from human judgments using probabilistic clustering and large language models. It builds a dataset of 300 scenarios involving six core actions and three moral rules, and employs an LLM-based preprocessing pipeline to cluster actions and scenarios. COMETH achieves roughly double the alignment with human judgments compared to end‑to‑end LLM prompting, while providing transparent explanations of the contextual features driving its predictions.
By Geoffroy Morlat, Marceau Nahon, Augustin Chartouny, Raja Chatila, Ismael T. Freire, Mehdi Khamassi
The paper introduces WSF-ARG+, a new dataset that pairs hate speech with check‑worthiness annotations, and presents an LLM‑in‑the‑loop framework to streamline the annotation process. Experiments with 12 open‑weight large language models demonstrate that the framework cuts human effort while maintaining annotation quality. The study also shows that incorporating check‑worthiness labels improves hate‑speech detection performance, boosting macro‑F1 scores for large models by up to 0.213 and averaging 0.154 across models.
By Nicol\'as Benjam\'in Ocampo, Tommaso Caselli, Davide Ceolin