نسخة أولية وصول مفتوح
Mind2Dialogue: Training Human-Aware Language Models by Simulating User Mental States
As language models become more capable, long-term collaboration in learning, reasoning, and decision-making calls for a deeper understanding of the people they serve. Yet training such human-aware language models faces a fundamental supervision gap because current datasets for LLM assistant training contain few if any …