AI 热点

Hao AI Lab 将 GPT-6 Astra + Codex 机器人智能体任务提速约 3 倍

UCSD Hao AI Lab更新于 11月20日 09:59

However, hiding the model's time behind motion doesn't make the robot react fast. Each action is still based on what the robot saw about 5 s earlier, so it need

AI 导读

Hao AI Lab 通过合并工具调用往返、缩短上下文并让模型思考与机器人运动并行,使 GPT-6 Astra + Codex 智能体在测试任务中提速约 3x。杯子任务从 11.1 min 缩短至 3.4 min,毛巾折叠从约 18 min 缩短至约 6 min,机器人运动时间占比从 24% 升至 84%。不过每个动作仍基于约 5 s 前的视觉信息,环境变化后的反应延迟约为 5 s;博客为 https://t.co/BIaiZ2P5ch,代码为 https://t.co/MvWxzfMXiK。

推荐理由

精选AI 评分 68

原文拆解了合并工具往返、压缩上下文与并行规划动作的工程方法,可迁移到降低真实机器人智能体的等待时间。

主题

标签

正文

However, hiding the model's time behind motion doesn't make the robot react fast. Each action is still based on what the robot saw about 5 s earlier, so it needs about 5 s to react when something changes.

We also treated Codex and Astra as a black box. With open models served for robots, the server and the harness can be designed together, leaving much more room to cut model time and be even faster.

Blog: https://t.co/BIaiZ2P5ch Code: https://t.co/MvWxzfMXiK

来源

UCSD Hao AI Lab