Unverified60% confidenceFactExact time
RT-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.
2
Sources
60%
Confidence
Long-term
Relevance
7/18/2026
First Seen
Sources
Related Claims
UnverifiedGoogle DeepMind的RT-2（2023年）将机器人动作词元化为离散token，附加到视觉-语言模型PaLI-X和PaLM-E的输出词汇表中，展现出零样本泛化能力79% similarVerifiedSpeculative decoding was formally proposed by Google DeepMind in 2023; overall throughput can be improved by 2–3 times, and the output quality is completely equivalent to that of the original large model.78% similarVerifiedGoogle DeepMind released Genie in early 2024, an 11-billion-parameter foundation world model capable of learning to generate interactive 2D virtual worlds from unlabeled internet gaming videos.75% similarVerifiedThe emergent ability phenomenon was systematically documented by the Google DeepMind research team in 2022.75% similarVerifiedThe speculative decoding paradigm was proposed almost simultaneously by teams at Google Brain and DeepMind in 2023.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/553554API
curl https://kongchang.com/api/v1/knowledge/claims/553554MCP
get_claim(id=553554)