Unverified50% confidenceFactExact time
RT-2的基座是280亿参数的PaLM-E模型，动作被编码为离散token序列，与语言token在同一自回归框架下生成
1
Sources
50%
Confidence
Long-term
Relevance
7/15/2026
First Seen
Sources
LLM是机器人智能的错误地基？具身AI的路线之争
Related Claims
UnverifiedRT-2（2023年）将55B参数的视觉-语言模型PaLM-E微调为可直接输出离散化动作token的策略模型76% similarUnverifiedLLM 0.32a0 introduced two core changes: modeling model input as a sequence of messages, and modeling model responses as typed streaming parts65% similarUnverifiedThe model deployment stage involves model quantization, inference acceleration frameworks (such as vLLM and Ollama), and API encapsulation.63% similarUnverifiedThe tool-calling mechanism allows large language models to dynamically invoke external functions during inference, representing a fundamental leap from language models to action models.62% similarUnverified即梦AI Agent模式的技术依赖于大语言模型（LLM）的推理能力与多模态生成模型的协同配合61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/514040API
curl https://kongchang.com/api/v1/knowledge/claims/514040MCP
get_claim(id=514040)