Proactive Dialogue Policy Optimization via Cognitive-State Transition
Across three tasks, Cog-Sim exhibits monotonic dose--response relationships and is preferred over prompt-based simulators for naturalness and CSTPO improves Qwen3-14B's performance to a level comparable to that of GPT-5.5-based planning methods.
Ming-Hui Ma, Meng-Qi Chen, Bin Guo et al.
· 0 citations