菜单
Methodology for evaluating autonomous AI capability progression across phases — measuring when models transition from human-assisted to autonomous execution using robotics tasks as a benchmark.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/hiyenwong/ai_collection --skill project-fetch-autonomous命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Graph-native Python reimplementation of the Information Dynamics of Music (IDyOM) model that represents predictive memories as explicit graph objects for musical expectation modeling and network analysis.
Physics-aware end-to-end deep reinforcement learning methodology for quadcopter control with actuator dynamics modeling.
Reinforced Dreamer methodology for asymmetric reinforcement learning using latent guidance to improve world model representations and behaviors in model-based RL.
基于 SOC 职业分类
正在显示 SKILL.md
| name | project-fetch-autonomous |
| description | Methodology for evaluating autonomous AI capability progression across phases — measuring when models transition from human-assisted to autonomous execution using robotics tasks as a benchmark. |
| category | ai_collection |
| trigger_words | project fetch, autonomous capability, robotics evaluation, capability progression, phase comparison, robodog, physical world agents |
| date | 2026-07-13T00:00:00.000Z |
| source | Anthropic Research - Project Fetch: Phase two (Jun 18, 2026) |
| url | https://www.anthropic.com/research/project-fetch-phase-two |
A methodology for evaluating how AI model capabilities progress from human-assisted to autonomous execution, using robotics tasks as a benchmark. The core pattern observed: first, models are helpful to humans. Then, humans are helpful to models. Finally, models are largely able to do things themselves.
| Capability Stage | Characteristics |
|---|---|
| Stage 1: Human-helpful | Model assists humans, humans make key decisions |
| Stage 2: Human-helpful-to-model | Humans guide model through difficult steps |
| Stage 3: Autonomous | Model completes tasks independently, 10-37x faster than human teams |
This methodology has been applied to:
The pattern suggests that domains following this progression will eventually see autonomous model capabilities, but NOT necessarily at the level of low-level physical control (e.g., developing specific actuation policies).
