Unverified50% confidenceFactExact time
DeepMind的Flamingo利用门控交叉注意力层将视觉特征注入冻结的语言模型中间层
1
Sources
50%
Confidence
Long-term
Relevance
9/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedDeepMind的Griffin将基于线性递归单元LRU的循环层与局部注意力层交替堆叠72% similarUnverified大语言模型从本质上是深度神经网络69% similarUnverifiedDeepMind提出RLAM（强化学习辅助记忆），OpenAI探索过程监督（Process Supervision）方法，将奖励信号延伸至中间推理步骤68% similarUnverifiedDeepMind开发了名为Genie的交互式世界模型系统68% similarVerifiedGoogle DeepMind的SIMA项目训练了一个能够在多款3D游戏中执行自然语言指令的通用智能体67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/854181API
curl https://kongchang.com/api/v1/knowledge/claims/854181MCP
get_claim(id=854181)