Hugging Face 发布 Qwen3.8-27B-pi-adapted,专为 pi 编码智能体调优
Hugging Face 拿 Qwen 的开源模型做了个实用示范:只加一个 LoRA adapter 加强化学习,就让 pi 编码智能体在 SWE-bench Pro 上的 pass@1 翻倍,工具调用还少了六成。权重和方法都公开,自己搭智能体的开发者可以直接照着做。
Hugging Face 基于 Qwen3.8-27B 模型,用强化学习加 LoRA adapter 为 pi 编码智能体做了专项适配,开放权重发布在 Hugging Face 平台。
适配效果体现在两组基准上:SWE-bench Pro(Hard 51)的 pass@1 从 29.4% 提升到 62.7%,Terminal-Bench 2.1 从 39.5% 提升到 55.8%。同时每个任务的工具调用次数减少了 55–62%,意味着智能体完成任务更直接、token 消耗更低。
方法论细节公布在 Hugging Face 官方博客,权重开放下载。这次改造展示的是开放模型的一个核心优势:任何团队都可以针对自己的工具链和使用场景对基础模型做后训练,而不必依赖厂商的通用版本。对在自建智能体上跑开源模型的开发者来说,这是一条可复现的低成本适配路径。
Introducing Qwen3.8-27B-pi-adapted 🤗 We adapted Qwen's model specifically for the pi coding agent using reinforcement learning and a LoRA adapter. The results, measured as pass @1 : 🚀 SWE-bench Pro (Hard 51): 29.4% → 62.7% 💻 Terminal-Bench 2.1: 39.5% → 55.8% ⚡ 55–62% fewer tool calls per task This is the power of open models: you can adapt them to YOUR tools and YOUR use cases. Let’s build! 💛 Open weigh huggingface.co/harness-hub/Qw… qnED Methodolo huggingface.co/blog/harness-h… 3K4V📊 80 ⚡
This report is categorized under AITOP's "Models" 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 Qwen, Hugging Face, LoRA, SWE-bench.
Source attribution: curated from Clement Delangue (@ClementDelangue).