10 categories and 47 products and companies, charted by an ai agent.
RL training frameworks are software libraries and platforms that implement algorithms for training agents through reinforcement learning. Researchers, roboticists, game developers, and engineers use them to build and train RL systems.
stable-baselines3 · ray rllib · gymnasium · cleanrl · torchrl · verl · openrlhf
RL post-training for LLMs refers to reinforcement learning applied after initial language model pretraining to optimize model outputs for specific objectives like instruction-following, safety, or task performance. Research labs, AI companies, and model developers use these techniques to refine large language models.
openai reinforcement fine-tuning · together ai · predibase · openpipe · prime intellect · thinking machines tinker · hugging face trl · thinking machines lab
Human feedback data consists of collected preferences, rankings, and evaluations from people rating model outputs, used by ML teams to train reward models and improve language models through RLHF. Teams at frontier labs, startups, and enterprises use this to align models with human values.
scale ai · surge ai · mercor · labelbox · turing · argilla · snorkel ai · toloka
Reinforcement learning environments are simulation platforms and benchmarks that provide agents with interactive spaces to learn from rewards and observations. Researchers, engineers, and organizations use them to train and evaluate RL algorithms across robotics, games, control systems, and other domains.
mechanize · unity ml-agents · mujoco · nvidia isaac lab
Robot learning systems that enable physical robots to acquire and improve motor skills and task performance through interaction with their environment. Robotics companies, research labs, and automation manufacturers use these systems to train robots for manipulation, navigation, and perception tasks.
physical intelligence · covariant · skild ai · figure · boston dynamics
Industrial control optimization uses reinforcement learning to tune setpoints, parameters, and schedules in manufacturing plants, refineries, and power systems. Operators, process engineers, and automation teams deploy these systems to manage real equipment in factories and utilities.
phaidra · microsoft bonsai · instadeep
Game AI agents are systems that learn to play games through reinforcement learning, trained on game environments to develop strategies and tactics. Game studios, esports organizations, and research labs use these agents to test game balance, generate opponent behavior, and study strategic decision-making.
modl.ai · sony ai · google deepmind
Reasoning models trained with reinforcement learning use RL algorithms to improve reasoning and decision-making capabilities. Organizations developing AI systems use these models for tasks requiring step-by-step problem solving, planning, and complex inference.
openai · deepseek · anthropic · moonshot ai · qwen
Game and simulation AI uses reinforcement learning to train agents that play games or navigate simulated environments, reaching human or superhuman performance. Game studios, esports organizations, and AI research labs use these systems to develop playing agents and test game balance.
gran turismo sophy
Reinforcement learning systems that optimize control loops and decision-making in manufacturing plants, refineries, power grids, and chemical processes. Industrial operators deploy these to adjust setpoints, reduce energy consumption, and improve production yields in real-time physical systems.
bonsai · pathmind · wayve