AI 精选动态
智能评分 63
Vida 开源 BrowserBC，复用浏览器会话技能
AI 推荐理由
查看原文获取完整的 benchmarks 数据和开源实现，评估技能复用对代理性能的提升核心解读
Vida 开源 BrowserBC 框架，将浏览器会话转化为可复用技能，实现 WebArena‑Hard 通过率从 60.5% 提升至 81.4%、ClawBench 通过率从 32.9% 提升至 68.4%，并将平均工具调用数从 31.2 次减少至 22.7 次。
全文
Vida open-sourced BrowserBC, a framework that allows users to turn browser sessions into reusable skills for AI agents.
> Instead of recalculating navigation on every turn, agents can follow a skill created from earlier task execution.
> Vida reports a substantially higher success rate with fewer steps, via the same AI agent.
Hotel booking bench? 👀
https://video.twimg.com/amplify_video/2071340680954159104/vid/avc1/1920x1080/qIsZSDeEVBs1pdQB.mp4?tag=28
> **引用原帖 Vida (@vida_agent):**
> We open-sourced BrowserBC:
> A system that turns human browser trajectories into reusable agent skills. Just one recording is enough to generalize a skill.
> 🛠️ GitHub: [https://t.co/WP8mQGuJ6N]
> Here’s how it works.
> 👇
> https://x.com/vida_agent/status/2070921732459024492
🚨 AI News | TestingCatalog (@testingcatalog): It is also possible to execute these skills with a cheaper model.
Overall benchmarks 👀
> WebArena-Hard overall from 60.5 to 81.4
> ClawBench from 32.9 to 68.4
> Mean tool calls drop from 31.2 to 22.7
BrowserBC is open-source and can be used locally with Claude Desktop and Claude Code.
Star the repo👇
https://t.co/uXtgiKSNTO