arXiv · NLP· Okenla AbdulQahhar Abimbola, Maurice Abraham John, Adepoju Jeremiah Moyondafoluwa, Omolaiye Samuel, Oko-Oba Sandra Chioma, Owoo Oluwamayowa Emmanuel, Thanni Adewuyi, Lily Aimas-Love, Benson Eniola, Abdulquddus Ajibade, Sylvia Essessien, Anyanwu Princess Emmanuella, Adereni Abiodun·· 3 天前AI 评分39
离线优先量化 AI:低连接环境下设备端母婴健康助手的设计与评估
Offline-First Quantized AI: Design and Evaluation of an On-Device Maternal-Health Assistant for Low-Connectivity Environments
AI 导读
该研究发布整合量化语言模型与本地知识库的离线 Android 母婴健康助手,在 6GB 手机上实现 14.3s 中位延迟。对比显示 Gemma 4 E2B 响应质量更高,150 条响应均分 3.80/4;Qwen3.5 0.8B 生成更快且内存占用更低。
来源:arXiv · NLP · arxiv.org