Computational Narratology
Beyond Hallucination: Emotion-Label Drift in LLMs and Objective Projection
Not every LLM failure is hallucination. Models state feelings instead of constructing them. Which constraints can a machine actually check? The data answers.