用对抗迭代法提升对话模拟器真实度，助力心理支持系统评估
DIAL: Direct Iterative Adversarial Learning for Realistic Multi-Turn Dialogue Simulation
- 通过生成器与判别器的对抗迭代，逐步优化用户模拟器
- 在心理支持场景中恢复被微调削弱的词汇多样性，判别准确率显著下降
- 模拟失败率与真实情况高度一致，适合高风险对话系统测试
真实用户模拟对多轮对话系统的训练与评估至关重要，但准确复现人类行为仍是重大挑战。本文提出直接迭代对抗学习（DIAL），通过生成器（用户模拟器）与判别器之间的竞争动态，迭代提升模拟器真实性。在心理健康支持领域，该方法有效恢复了因监督微调而丧失的词汇多样性，并显著降低判别器准确率。模拟结果与真实场景的失败发生率高度相关，同时保持失败模式分布差异小。DIAL为多轮对话系统开发真实用户模拟器提供了可靠且低成本的解决方案，适用于部署前的系统评估。
原文摘要 · Abstract (English)
Realistic user simulation is crucial for training and evaluating multi-turn dialogue systems, yet creating simulators that accurately replicate human behavior remains a significant challenge. An effective simulator must expose the failure modes of the systems under evaluation. This work introduces Direct Iterative Adversarial Learning (DIAL), an adversarial framework that iteratively enhances user simulator realism through a competitive dynamic between a generator (user simulator) and a discriminator. When applied to mental health support, a domain characterized by diverse failure types and a critical dependence on realistic user behavior for failure detection, DIAL restores lexical diversity diminished by supervised fine-tuning and drastically reduces discriminator accuracy. The resulting simulator exhibits a strong correlation between simulated and real failure occurrence rates while maintaining low distributional divergence of failure modes. These findings indicate that DIAL is a promising method for developing realistic user simulators in multi-turn dialogue, facilitating reliable and cost-effective system evaluation prior to deployment.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。