Unverified50% confidenceFactExact time
SayCan由Google DeepMind于2022年提出，将语言模型与机器人动作库结合
1
Sources
50%
Confidence
Long-term
Relevance
7/16/2026
First Seen
Sources
Related Claims
UnverifiedGoogle于2022年发布SayCan，让语言模型仅负责生成可行动作序列，具体执行交由预训练底层技能策略完成77% similarUnverified将大语言模型接入机器人的典型范式包括Google的SayCan（用LLM分解任务后调用底层运动原语）和RT-2（端到端生成控制指令）74% similarUnverifiedRT-2, released by Google DeepMind in 2023, was the first to validate the feasibility of fine-tuning robot control on a 55-billion-parameter vision-language model, with generalization capability nearly doubling compared to its predecessor.71% similarUnverifiedThe ReAct framework was originally proposed by Google DeepMind researchers in 2022, with its core being the alternating execution of language model reasoning and external tool calls.70% similarUnverified工具增强语言模型（Tool-Augmented Language Model，TALM）最早的系统性论述可追溯至 2022 年 Google 发表的同名论文67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/533815API
curl https://kongchang.com/api/v1/knowledge/claims/533815MCP
get_claim(id=533815)