01Entry
Mind2Dialogue: Training Human-Aware Language Models by Simulating User Mental States
Trains LMs to simulate user mental states before responding — 'understand the person behind the prompt' as a trainable objective, not a prompt-engineering trick. Relevant to agent UX and goal inference over literal instruction-following.