Skywork(@Skywork_ai)
What changed under the hood: → Trained on a self-built OpenClaw environment, with tools and tasks s...
6.5Score
TL;DR · AI 摘要
Skywork 的新模型在自建的 OpenClaw 环境中训练，性能优于前沿开源模型，并提供两种变体。
核心要点
- Skywork 模型在自建 OpenClaw 环境中训练。
- 模型在 PinchBench、Claw-Eval 和 Skywork-Claw-Bench 上表现优异。
- 提供 v1.0 和 v1.0-lite 两种变体。
结构提纲
按章节快速跳转。
思维导图
用一张图看清主题之间的关系。
查看大纲文本（无障碍 / 无 JS 友好）
- Skywork 新模型
金句 / Highlights
值得收藏与分享的关键句。
模型在自建的 OpenClaw 环境中训练。
模型在 PinchBench、Claw-Eval 和 Skywork-Claw-Bench 上表现优异。
提供 v1.0 和 v1.0-lite 两种变体。
#Skywork#OpenClaw#模型训练#性能评估
打开原文→ Trained on a self-built OpenClaw environment, with tools and tasks synthesized from real user patterns. → Outperforms frontier open models on PinchBench, Claw-Eval, and Skywork-Claw-Bench. → Two variants: v1.0 (full) and v1.0-lite (faster," / X
Skywork on X: "What changed under the hood: → Trained on a self-built OpenClaw environment, with tools and tasks synthesized from real user patterns. → Outperforms frontier open models on PinchBench, Claw-Eval, and Skywork-Claw-Bench. → Two variants: v1.0 (full) and v1.0-lite (faster," / X
Don’t miss what’s happening
What changed under the hood: → Trained on a self-built OpenClaw environment, with tools and tasks synthesized from real user patterns. → Outperforms frontier open models on PinchBench, Claw-Eval, and Skywork-Claw-Bench. → Two variants: v1.0 (full) and v1.0-lite (faster, lower cost).
·
1
2