微软发布FrogNano小模型编程智能体
微软教你用合成任务训练小模型编程智能体，不依赖大模型蒸馏，在普通设备上就能运行。
微软研究团队发布FrogNano，一个40亿参数的编程智能体，在1500个软件工程环境中训练。该模型不依赖传统蒸馏技术，完全通过强化学习在合成任务上训练。
研究团队采用在线任务合成方法，为模型生成恰好处于其学习能力边界的问题。
Super cool paper from Microsoft. They show that it's possible to build competitive small coding agents without traditional distillation from frontier models. DAIR.AI @dair_ai Banger report from Microsoft. (bookmark it) They show that it's possible to build competitive small coding agents without traditional distillation from frontier models. This is a big deal! The work describes how they achieved this. They introduce a 4B coding agent trained on roughly 1,500 software engineering environments. The cool thing is that they use no distillation from a larger model at any point. FrogNano is post-trained purely with RL on synthetic tasks. The target is a coding agent that runs on minimal machines, which rules out both a frontier backbone and a frontier teacher. The ingredient the report credits the most is online task synthesis. The pipeline generates tasks calibrated to the frontier of learnability for the current checkpoint, so the agent always trains on problems it can just barely solve. The authors argue that calibration, rather than the volume of synthetic data, is what makes this work. This means that competitive small coding agents can be trained from synthetic tasks alone. And generating those tasks at the current agent's learnability frontier is what makes this particular training productive. The report covers training methodology, evaluations across diverse environments, and analyses of what the agent learned. Paper: academy.dair.ai/papers/frognan… 🔗 View Quoted Tweet📊 7 ⚡
This report is categorized under AITOP's "Papers" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include FrogNano, Microsoft, 编程智能体, 强化学习.
Source attribution: curated from elvis (@omarsar0).