通过 Reason-of-Select 蒸馏实现持续对话状态跟踪
Continual Dialogue State Tracking via Reason-of-Select Distillation
AI总结：
针对对话状态跟踪中的灾难性遗忘和“Value Selection Quandary”问题，提出Reason-of-Select蒸馏方法，通过增强小模型的元推理能力、领域引导及两项改进策略，有效缓解遗忘并实现可靠知识转移，实验证明其性能与泛化能力卓越。
摘要：
理想的对话系统需要在保留先验知识的同时，持续获取技能并适应新任务。在这些系统中至关重要的对话状态跟踪（DST），通常涉及学习新服务并面临灾难性遗忘，以及被称为“Value Selection Quandary”的关键能力丧失。为了应对这些挑战，我们通过增强较小模型的新颖“元推理”能力，引入了 Reason-of-Select (RoS) 蒸馏方法。元推理采用增强的多领域视角，在持续学习期间结合来自特定领域对话的元知识片段。这超越了传统的单一视角推理。领域引导过程增强了模型从多个可能值中剖析复杂对话的能力。其领域无关属性对齐了不同领域的数据分布，有效缓解了遗忘。此外，两项新颖的改进——“多值解析”策略和语义对比推理选择方法，通过生成 DST 特定的选择链并缓解教师推理中的幻觉，显著增强了 RoS，确保了有效且可靠的知识转移。大量实验验证了我们方法的卓越性能和强大的泛化能力。提供了源代码以确保可复现性。
英文摘要：
An ideal dialogue system requires continuous skill acquisition and adaptation to new tasks while retaining prior knowledge. Dialogue State Tracking (DST), vital in these systems, often involves learning new services and confronting catastrophic forgetting, along with a critical capability loss termed the "Value Selection Quandary." To address these challenges, we introduce the Reason-of-Select (RoS) distillation method by enhancing smaller models with a novel 'meta-reasoning' capability. Meta-reasoning employs an enhanced multi-domain perspective, combining fragments of meta-knowledge from domain-specific dialogues during continual learning. This transcends traditional single-perspective reasoning. The domain bootstrapping process enhances the model's ability to dissect intricate dialogues from multiple possible values. Its domain-agnostic property aligns data distribution across different domains, effectively mitigating forgetting. Additionally, two novel improvements, "multi-value resolution" strategy and Semantic Contrastive Reasoning Selection method, significantly enhance RoS by generating DST-specific selection chains and mitigating hallucinations in teachers' reasoning, ensuring effective and reliable knowledge transfer. Extensive experiments validate the exceptional performance and robust generalization capabilities of our method. The source code is provided for reproducibility.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
