Reposted by Yoav Ger
Are world models necessary to achieve human-level agents, or is there a model-free short-cut?
Our new #ICML2025 paper tackles this question from first principles, and finds a surprising answer, agents _are_ world models… 🧵
arxiv.org/abs/2506.01622
Our new #ICML2025 paper tackles this question from first principles, and finds a surprising answer, agents _are_ world models… 🧵
arxiv.org/abs/2506.01622
June 4, 2025 at 3:48 PM
Are world models necessary to achieve human-level agents, or is there a model-free short-cut?
Our new #ICML2025 paper tackles this question from first principles, and finds a surprising answer, agents _are_ world models… 🧵
arxiv.org/abs/2506.01622
Our new #ICML2025 paper tackles this question from first principles, and finds a surprising answer, agents _are_ world models… 🧵
arxiv.org/abs/2506.01622
Reposted by Yoav Ger
All our motor control modelling efforts focus on closed-loop systems for this reason:
"...closed-loop and open-loop training produce fundamentally different learning dynamics, even when using identical architectures and converging to the same final solution."
arxiv.org/abs/2505.13567
"...closed-loop and open-loop training produce fundamentally different learning dynamics, even when using identical architectures and converging to the same final solution."
arxiv.org/abs/2505.13567
Learning Dynamics of RNNs in Closed-Loop Environments
Recurrent neural networks (RNNs) trained on neuroscience-inspired tasks offer powerful models of brain computation. However, typical training paradigms rely on open-loop, supervised settings, whereas ...
arxiv.org
May 29, 2025 at 5:28 PM
All our motor control modelling efforts focus on closed-loop systems for this reason:
"...closed-loop and open-loop training produce fundamentally different learning dynamics, even when using identical architectures and converging to the same final solution."
arxiv.org/abs/2505.13567
"...closed-loop and open-loop training produce fundamentally different learning dynamics, even when using identical architectures and converging to the same final solution."
arxiv.org/abs/2505.13567