# Introduction The world of artificial intelligence is taking a significant turn in its ability to understand and evaluate moral actions. A new platform, COMETH (Contextual Organization of Moral Evaluation from Textual Human inputs), is revolutionizing the learning of Llama models. COMETH is a framework that integrates a probabilistic contextual learner with LLM-based semantics and human moral evaluations to model how context influences the perception of moral ambiguity in actions. The platform cares for an empirically grounded dataset of 300 scenarios on six key actions (violate not kill, do not deceive, and do not break the law) and collects ternary judgments (Blame/Neutral/Support) from N=101 participants. A standard pre-processing pipeline standardizes actions through a LLM filter and miniLM with K-means, producing robust and reproducible core-action clusters. COMETH learns specific moral contexts for each action through online clustering of human scenarios using base distribution judgments and principal divergence criteria. A generalization module extracts concise contextual features and learns weights in a transparent probability-based model. Empirically, COMETH doubles the agreement with human majority judgments compared to end-to-end LLM solutions (around 60% vs around 30% on average), while revealing which contextual features guide its predictions. # Contributions COMETH presents three significant contributions: (i) an empirically grounded moral-context dataset, (ii) a reproducible pipeline combining human evaluations with context-based learning and LLM semantics, and (iii) an interpretable alternative to end-to-end LLMs for sensitive context-dependent predictions and explanations. ## Conclusion COMETH represents a significant turning point in understanding moral intelligence. Its ability to model how context influences the perception of moral actions could have a profound impact on creating more responsible and ethical systems. ## Implications The COMETH platform opens up new perspectives for learning Llama models and may be used to improve moral understanding in various fields, from medicine to finance.