[KongchangAI]
Concept循环架构 / 循环推理模型

Recurrent Models

A class of neural network architectures that extend effective depth by repeatedly executing the same module to accomplish multi-step reasoning, iteratively updating a state vector in latent space.

Timeline (last 90 days)

Nov 13

Recurrent models perform well on simple reasoning puzzles (such as Sudoku and mazes), but training is unstable and slow, and parameter and computation overhead is large.

Unverified50%
Nov 13

On Sudoku and maze tasks, CRAM trains at least 1000 times faster than recurrent models.

Unverified50%
Nov 13

On Sudoku and maze tasks, CRAM reduces inference FLOPs by at least 282 times compared with recurrent models.

Unverified50%
Nov 9

Recurrent models lack an explicit measure of progress, so stopping conditions can only be determined by external rules or a fixed number of iterations, and the model cannot assess the quality of intermediate states.

Unverified50%
Nov 9

If energy measures can be combined with more complex search strategies in the future, the potential of recurrent models on tasks such as combinatorial optimization and logical reasoning is worth anticipating.

Unverified50%

All Facts (5)

Source Articles