标签: ChatGPT

A Princeton Researcher Proposes Recurrent Looped Transformer (RLT) that Carries Decoder State across Every Token, Fixing 96 Blocks per Token with Unbounded Temporal Depth

A Princeton Researcher Proposes Recurrent Looped Transformer (RLT) that Carries Decoder State across Every Token, Fixing 96 Blocks per Token with Unbounded Temporal Depth

普林斯顿研究者提出 Recurrent Looped Transformer(RLT),把解码器状态在生成每个 token 时持续携带,并对每个 token 固定执行 96 个模块的循环计算,用“无界时间深度”替代单纯堆叠层数。它值得关注,因为这是在挑战大模型扩容的一个基础假设:性能提升是否只能靠更多参数和更…