Lightweight Adaptation for LLM-based Technical Service Agent: Latent Logic Augmentation and Robust Noise Reduction
轻量级适应用于基于LLM的技术服务代理:潜在逻辑增强与鲁棒噪声消除
机构 * School of Mathematics and Sciences, Fudan University, Shanghai, China(复旦大学数学科学学院,上海,中国) ; Shanghai Center for Mathematical Sciences, Fudan University, Shanghai, China(复旦大学上海数学中心,上海,中国) ; Alibaba Group, Hangzhou, China(阿里巴巴集团,杭州,中国) ; Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University, Shanghai, China(脑启发式智能科学技术研究院,复旦大学,上海,中国) ; Center for Applied Mathematics & Shanghai Key Laboratory of Contemporary Applied Mathematics, Fudan University, Shanghai, China(应用数学中心及上海当代应用数学重点实验室,复旦大学,上海,中国)
AI总结 本文提出轻量级适应框架,通过潜在逻辑增强和鲁棒噪声消除提升技术服务代理的稳定性和性能,同时采用混合奖励机制降低训练成本。