1月12日，梅赛德斯-奔驰宣布全新奔驰纯电CLA虚拟助手迎来升级，新增多项功能。 奔驰虚拟助手依托豆包大模型的AI生成能力，融合思必驰全链路语音交互技术与中枢大模型架 构，实现高效语义理解与情感化话术生成，于2025年底随车型上市并持续优化。 此次奔驰虚拟助手将优化短期对话记忆和深度语义关联——唤醒时间仅需0.2秒，端到端执行时间 缩短至0.8秒，实现近乎即时的反馈。多音区交互识别定位准确率可以达到98%，并支持连续对话和自 然语言处理，系统可以保持上下文关联，准确理解指代关系。同时，优化了短期对话记忆和深度语义关 联能力，使多轮交互更顺畅。 另外，新增奔驰用车指南和拟人对话模型，强化了情感化体验。通过情绪识别技术，系统可识别用 户情绪状态，并通过Avatar形象色彩变化、TTS语音播报内容调整及环境音效联动提供个性化回应。 ...

Core Viewpoint - The collaboration between SIBIC and Audi has successfully transitioned into mass production, focusing on enhancing in-car voice interaction technology for localized models based on Audi's PPE and PPC platforms [1][8]. Group 1: Voice Interaction Technology - SIBIC has developed a comprehensive voice interaction system for Audi, optimizing wake-up response, dialogue Q&A, and command execution to improve interaction efficiency and user experience [3]. - The system supports rapid wake-up, achieving a 500ms first character display and 1300ms end-to-end interaction, allowing users to issue subsequent commands immediately after wake-up [5]. - The system can understand and execute complex commands containing multiple intents in a single sentence, enhancing user convenience [5]. Group 2: Offline Functionality and Customization - SIBIC's technology includes offline voice recognition and semantic understanding, ensuring stable performance in weak or no network environments, such as outdoors or in basements [6]. - The system adapts to user speech patterns and dialogue contexts, dynamically adjusting voice segmentation intervals for a more natural conversation flow [6]. - Users can create personalized voice profiles for the voice assistant through a simple recording process, allowing for customized voice interactions [6]. Group 3: Multilingual Support and Strategic Development - The system supports Mandarin with regional accents, Cantonese, and English, catering to diverse user needs across different regions [6]. - Audi is advancing its "PPC+PPE" dual-platform strategy, focusing on the intelligent development of both fuel and electric vehicles, with the PPC platform covering B to D-class models [6][8]. - The collaboration aims to enhance the voice interaction experience for various localized models, with plans for deeper cooperation to evolve voice commands into proactive, scenario-based services [8].