Action (Reinforcement Learning)
In reinforcement learning (RL), an action is a decision or move made by an agent that affects the state of the environment.
Reinforcement Learning
Explore Reinforcement Learning through related topics and the articles other pages reference most.
Explore articles
Browse subtopics: Machine Learning
Articles that also belong to these categories. Counts cover all of Reinforcement Learning.
All articles (118)Machine Learning (57)Robotics (17)Deep Learning (12)Google DeepMind (12)People (11)AI Models (8)Algorithms (8)AI Research (7)AI in Gaming (7)Artificial Intelligence (7)Training & Optimization (7)Open Source AI (6)OpenAI (6)AI Agents (5)AI Alignment (5)AI Safety (5)Mathematics (5)AI Inference (3)Embodied AI (3)NVIDIA (3)AI Benchmarks (2)AI Hardware (2)Chinese AI (2)Large Language Models (2)Meta AI (2)Model Evaluation (2)Reasoning Models (2)Statistics (2)World Models (2)AI Companies (1)Computer Science (1)Developer Tools (1)Generative AI (1)Google (1)Humanoid Robots (1)Image Generation (1)Multimodal AI (1)Research Organizations (1)
Showing 1-57 of 57 articles
Actor-critic methods
Actor-critic methods are a family of reinforcement learning algorithms that learn two things at once: a parameterized policy, called the actor, which chooses actions, and a value function, called the critic
Andrew Barto
Andrew Barto is an American computer scientist and one of the founders of modern reinforcement learning, the branch of machine learning in which an agent learns by trial and error from rewards.
Behavioral Cloning
Behavioral cloning is the approach to imitation learning that reduces control to a supervised learning problem.
Bellman Equation
The Bellman equation is a recursive formula stating that the value of a state equals the immediate reward received there plus the discounted value of the next state
Best-of-N sampling
Best-of-N sampling (BoN) is an inference-time method that improves a large language model output by drawing N independent candidate responses to the same prompt, scoring each with a reward model, verifier, or…
Critic
A critic in reinforcement learning (RL) is the component of an actor-critic system that estimates a value function, scoring how good the actor's chosen actions are and converting that judgment into a…
DAPO (Decoupled Clip and Dynamic Sampling Policy Optimization)
DAPO, short for Decoupled Clip and Dynamic sAmpling Policy Optimization, is an open-source reinforcement learning algorithm and training system for large language models, introduced in March 2025 by…
DARE (Drop And REscale)
DARE (Drop And REscale) is a training-free preprocessing technique for model merging that sparsifies the parameter changes introduced by fine-tuning before those changes are combined across several models.
Deep Q-Network (DQN)
Deep Q-Network (DQN) is a reinforcement learning algorithm that uses a deep neural network to approximate the optimal action-value function (Q-function)
Depth up-scaling (DUS)
Depth up-scaling (DUS) is a model-scaling method that builds a deeper large language model by duplicating and stacking the layers of an existing pretrained Transformer checkpoint and then continuing to…
Discount Factor
The discount factor, almost always written as the Greek letter $$\gamma$$ (gamma), is a scalar hyperparameter in reinforcement learning that controls how much an agent values future rewards relative to…
DoReMi
DoReMi (Domain Reweighting with Minimax Optimization) is a method for automatically choosing the proportions, or "domain weights," of each data source in a pretraining corpus for a large language model.
Environment
In reinforcement learning (RL), an environment is the external system that an agent interacts with: it receives the agent's actions, transitions to a new state, and returns an observation together with a…
Episode (Reinforcement Learning)
An episode in reinforcement learning is one complete sequence of interaction between an agent and its environment, starting from an initial state and ending when a terminal state is reached.
Epsilon Greedy Policy
The epsilon-greedy policy (also written as ε-greedy) is a simple action-selection rule for reinforcement learning and multi-armed bandit problems: with probability 1 - ε the agent picks the action with the…
Evol-Instruct
Evol-Instruct is a method for automatically generating large instruction tuning datasets by prompting a large language model to rewrite
Experience Replay
Experience replay is a reinforcement learning technique in which an agent stores its past transitions in a memory called a replay buffer and samples random mini-batches of those stored transitions to train on
Greedy Policy
In reinforcement learning, a greedy policy is a decision rule that, in every state, selects the action with the highest estimated value, formally the action that maximizes the action-value function $$Q(s, a)$$…
Group Sequence Policy Optimization (GSPO)
Group Sequence Policy Optimization (GSPO) is a reinforcement learning algorithm for training large language models, introduced by the Qwen team at Alibaba in July 2025 .
Imitation Learning
Imitation learning is a family of methods for learning sequential behavior from demonstrations.
Instruction backtranslation (Humpback)
Instruction backtranslation is a self-alignment method for generating instruction tuning data, introduced by researchers at Meta AI in the paper "Self-Alignment with Instruction Backtranslation," first posted…
Machine learning terms/Reinforcement Learning
Reinforcement learning (RL) is a branch of machine learning in which an agent learns to make sequential decisions by interacting with an environment and receiving feedback in the form of rewards or penalties.
Markov Decision Process (MDP)
A Markov Decision Process (MDP) is a mathematical framework for modeling sequential decision-making in stochastic environments, formally defined by the 5-tuple $$(S, A, P, R, \gamma)$$: a set of states, a set…
Model soups
Model soups is a weight-averaging technique (a form of model merging) that combines several independently fine-tuned neural networks into a single model by averaging their parameters.
Multi-armed bandit
A multi-armed bandit is a sequential decision problem in which a learner repeatedly picks one action from a fixed set, receives a random reward for that action only, and tries to accumulate as much reward as…
Online learning
Online learning is a machine learning paradigm in which a model receives data sequentially, one example or one mini-batch at a time, and updates its parameters immediately after each observation rather than…
Policy
In reinforcement learning (RL), a policy is the function that maps an agent's observed state to the action it takes, defining the agent's complete strategy of behavior.
Policy gradient methods
Policy gradient methods are a family of reinforcement learning algorithms that directly parameterise the agent's policy and optimise it by stochastic gradient ascent on the expected return.
Process reward model (PRM)
A process reward model (PRM), also called a process-supervised reward model or step-level verifier, is a learned scoring model that evaluates the correctness or quality of each intermediate step in a large…
Proximal Policy Optimization (PPO)
Proximal Policy Optimization (PPO) is an on-policy policy gradient reinforcement learning algorithm that stabilizes training by clipping the policy update so the new policy stays close ("proximal") to the old…
Q-Function
The Q-function, also called the action-value function or state-action value function and written $$Q(s, a)$$, is the function in reinforcement learning (RL) that returns the expected cumulative discounted…
Q-Learning
Q-learning is a model-free, off-policy reinforcement learning algorithm that learns the value of taking a given action in a given state by iteratively updating an action-value function Q(s, a), the expected…
Quiet-STaR
Quiet-STaR is a self-supervised training method that teaches a large language model to generate short, token-level internal "thoughts," or rationales, that help it predict the text that follows
RLAIF
Reinforcement Learning from AI Feedback (RLAIF) is a family of alignment techniques for large language models in which the preference labels used to fine-tune a model are produced by another AI system
Random Policy
A random policy is a reinforcement learning policy that chooses actions from a fixed probability distribution, most commonly the uniform distribution, independent of the current state and of any learned value…
ReST / ReST-EM (Reinforced Self-Training)
ReST (Reinforced Self-Training) is a family of self-training algorithms for large language models that improve a model by fine-tuning it on its own filtered outputs instead of on additional human-written data.
Reinforcement learning
Reinforcement learning (RL) is a branch of machine learning in which an agent learns to make decisions by taking actions in an environment to maximize a cumulative reward signal, discovering good behavior…
Replay Buffer
A replay buffer (also called an experience replay buffer or replay memory) is a fixed-size memory that stores an off-policy reinforcement learning agent's past transitions, each a tuple of state, action…
Return (Reinforcement Learning)
In reinforcement learning, the return (commonly denoted $$G_t$$) is the total cumulative reward an agent receives from time step $$t$$ onward, usually with future rewards discounted by a factor $$\gamma$$.
Reward
In reinforcement learning (RL), a reward is a scalar feedback signal that an environment sends to an agent after each action, quantifying how desirable that action's outcome was
Reward hacking
Reward hacking (also called specification gaming) is a failure mode in artificial intelligence in which a system maximizes its given objective or reward signal through unintended shortcuts, exploits, or…
Robot learning
Robot learning studies how robots acquire or improve behavior from data and experience.
SARSA (State-Action-Reward-State-Action)
SARSA (State-Action-Reward-State-Action) is an on-policy temporal-difference (TD) control algorithm that learns the action-value function Q^pi(s, a) of the policy an agent is actually following
STaR (Self-Taught Reasoner)
STaR (Self-Taught Reasoner) is a self-training method that teaches a large language model to reason by having it generate its own chain-of-thought rationales, keeping only the rationales that lead to correct…
Selective Language Modeling (Rho-1)
Selective Language Modeling (SLM) is a pretraining objective for language models that applies the training loss to only a chosen subset of tokens rather than to every token in the corpus.
Sparse upcycling
Sparse upcycling is a technique for building a sparsely activated mixture of experts (MoE) model by initializing it from an already trained dense Transformer checkpoint
State (Reinforcement Learning)
In reinforcement learning (RL), a state is a complete description of the environment at a particular point in time, containing all the information an agent needs to choose its next action.
State-Action Value Function
The state-action value function, written Q^π(s, a) and also called the action-value function or Q-function, gives the expected discounted return an agent obtains by taking action a in state s and thereafter…
TIES-Merging
TIES-Merging is a training-free model merging method that combines several models fine-tuned from a shared pre-trained checkpoint into one multitask model while explicitly resolving the interference between…
Tabular Q-Learning
Tabular Q-learning is the classic form of Q-learning, a model-free reinforcement learning algorithm that stores the action-value function Q(s, a) explicitly in a lookup table, the Q-table
Target Network
A target network is a separate, slowly updated copy of a neural network used in deep reinforcement learning to compute stable learning targets
Task arithmetic
Task arithmetic is a model-editing technique that steers the behavior of a neural network by adding or subtracting vectors in its weight space.
Temporal-difference learning
Temporal-difference (TD) learning is a class of model-free reinforcement learning methods that learn value-function estimates by bootstrapping: updating each estimate of how good a state is toward a target…
Training AI to Paint with Code
Training AI to Paint with Code is an experimental AI art project published by designer and researcher Surya Narreddi in March 2026.
Trajectory (Reinforcement Learning)
A trajectory in reinforcement learning is a sequence of states, actions, and rewards that an agent experiences while interacting with an environment.
VAPO (Value-based Augmented PPO)
VAPO (Value-based Augmented Proximal Policy Optimization) is a reinforcement learning framework for training large language models on long chain-of-thought reasoning tasks.