From Stateless to Situated: Building a Psychological World for LLM-Based Emotional Support

ArXi:2603.25031v1 Announce Type: new In psychological and emotional companionship scenarios, the core limitation of large language models (LLMs) lies not merely in response quality, but in their reliance on local next-token prediction, which prevents them from maintaining the temporal continuity, stage awareness, and user consent boundaries required for multi-turn intervention. This stateless characteristic makes systems prone to premature advancement, stage misalignment, and boundary violations in continuous dialogue.