Liquid AI 分享端侧小模型训练秘籍：1B 参数搞定手机 Agent
maximelabonne · x · 2026-08-03
Liquid AI's head of post-training detailed the full tech stack for building high-performance Small Language Models (SLMs). The presentation explains how to train a model under 1GB that runs efficiently on-device in 20 minutes.
The core training stack includes:
- LFM2.5 as the base model
- On-policy preference alignment (DPO)
- Agentic reinforcement learning
- Curriculum training combined with iterative model merging
Through this loop, frontier small models can now outperform traditional 70B models on critical agentic tasks like tool calling.
所属事件:Liquid AI 分享手机端 1B 小模型训练栈(2 条相关)→
「模型」频道最新
- 阿里开源 Qwen3.8-Max:2.4 万亿参数,专攻长周期复杂任务 — The Decoder · 2026-08-03
- 开发者实测推荐 KAT Coder 2.5：比 Qwen 更快更准 — The_Paradoxy · 2026-08-03
- 智谱 GLM-5.3 模型踪迹在代码库中被发现 — Few_Painter_5588 · 2026-08-03
- 瑜伽工作室聊天机器人架构：多模型路由与低成本记忆方案 — omi0009 · 2026-08-03
- 观点：Fable 5与GPT-5.6 Sol在长程数学推理上实现质的飞跃 — abeirami · 2026-08-03
- AI9Stars 发布开源模型 G9v3-39A5B，主打推理与智能体工作流 — Tall-Ad-7742 · 2026-08-03