Hierarchical reinforcement learning (HRL)
Hierarchical reinforcement learning (HRL) is a framework that structures learning tasks into a hierarchy, improving the efficiency of learning in complex environments.
reinforcement learning
Offline reinforcement learning
Offline reinforcement learning refers to a branch of machine learning where an agent learns from a fixed dataset without direct interaction with the environment.
Intrinsic motivation in AI
Intrinsic motivation in AI refers to the internal processes that drive artificial intelligence systems to pursue goals without external incentives.
Proximal policy optimization (PPO) for language models
Proximal policy optimization (PPO) is a reinforcement learning algorithm adapted for training language models by optimizing policies that generate text. It enables efficient and stable fine-tuning of language models with human feedback or reward signals.
Causal reinforcement learning
Causal reinforcement learning integrates causal inference with reinforcement learning, enabling agents to make decisions based on cause-and-effect relationships.
Empowerment (machine learning)
Empowerment in machine learning is an information-theoretic measure used to quantify the agent’s potential influence over its environment. It is applied primarily in reinforcement learning as an intrinsic motivation to encourage exploration and autonomous behavior.
Offline-to-online RL fine-tuning
Offline-to-online RL fine-tuning refers to the process of enhancing reinforcement learning models trained on offline data by further training them online.
Transformers for reinforcement learning (Decision Transformer)
The Decision Transformer is a model that applies transformer architectures to reinforcement learning, merging sequence modeling with decision-making tasks.
Latent world models (LWM)
Latent world models (LWM) are a framework in artificial intelligence that capture complex environments and dynamics for decision-making tasks.
A3C (asynchronous advantage actor-critic)
A3C is a reinforcement learning algorithm that utilizes asynchronous training to improve performance and efficiency in decision-making tasks.