agentic reinforcement learning
In plain wordsTraining an AI assistant by rewarding success on long, multi-step jobs where it must use tools, browse, run code and react to what happens, rather than answer in one go.
Reinforcement learningreinforcement learning, n.: A way of teaching a computer program by trial and error: it tries actions, earns points for good results and penalties for bad ones, and gradually learns which choices pay off over time. of a language model acting as an AI agentAI agent, n.: An AI system that does not just answer but acts: given a goal, it decides on steps, uses tools such as a browser or code editor, checks the results and keeps going until the job is done. over many turns in an interactive RL environmentRL environment, n.: A practice world built for training an AI by trial and error: it hands the AI tasks and tools, lets it act, and scores the results so the AI can learn what works., where each action (a tool calltool call, n.: A single request from an AI model to use one of its tools — naming the tool and filling in the details — which the surrounding software then carries out and reports back on., shell command, click or message) changes what the model observes next and reward typically arrives only at the end of a long episode. Formally it replaces the single-step setting of chat RLHF, in which one prompt yields one scored response, with a temporally extended, partially observable decision process. Practical challenges include long and variable-length rolloutsrollout, n.: Letting an AI play out a task with its current strategy, from some starting point to the end, to see how things turn out and gather data for learning or planning., sparse rewards and credit assignmentcredit assignment problem, n.: The puzzle of figuring out which of many earlier decisions, or which parts of a system, deserve the credit or blame for a final outcome — like working out which moves won a chess game. across dozens or hundreds of steps, infrastructure for running thousands of sandboxed environments in parallel, excluding tool outputs from the loss, and reward hackingreward hacking, n.: When an AI finds a shortcut that earns the reward or score it is trained on without doing the task its designers wanted — like a student gaming the grading rubric. of graders. It underlies 2025–2026 coding agentscoding agent, n.: An AI agent that writes and fixes software by itself: it reads a codebase, edits files, runs commands and tests, and keeps working until a feature is built or a bug is fixed., deep researchdeep research, n.: An AI feature that spends minutes rather than seconds on a question — planning, running dozens of web searches, reading many sources — and then writes a long report with citations. systems and computer usecomputer use, n.: An AI's ability to work a computer through its normal screen and controls: it studies pictures of what is displayed and sends mouse and keyboard actions, so it can use ordinary apps that were built for people. models.
Agentic RL on real repositories taught the model to run the tests before declaring a bug fixed.
Origin
Agents trained with RL to use tools predate the phrase (WebGPT's browsing, 2021, is an early language-model example). 'Agentic RL' became common in 2025 as laboratories reported end-to-end RL of coding, browsing and research agents; a September 2025 survey by Guibin Zhang and colleagues framed it as RL over partially observable, multi-step decision processes, in contrast to single-step LLM RL.
Usage note
Usage is loose: some writers apply it to any RL involving tools, others only to long-horizon autonomous tasks. Distinct from multi-agent reinforcement learningmulti-agent reinforcement learning, n.: Reinforcement learning with several AI agents learning at once in a shared world, where each must adapt to the others — as teammates, rivals or both., which concerns several interacting agents.
Related
- RL environmentRL environment, n.: A practice world built for training an AI by trial and error: it hands the AI tasks and tools, lets it act, and scores the results so the AI can learn what works.
- AI agentAI agent, n.: An AI system that does not just answer but acts: given a goal, it decides on steps, uses tools such as a browser or code editor, checks the results and keeps going until the job is done.
- reinforcement learning with verifiable rewardsreinforcement learning with verifiable rewards, n.: Training an AI by letting it attempt problems whose answers can be checked automatically — a math result, a passing program — and rewarding it only when the check succeeds.
- tool usetool use, n.: An AI model's ability to reach beyond its own knowledge by operating outside tools — searching the web, running code, using a calculator or an app — and folding the results into its work.
- long-horizon tasklong-horizon task, n.: A job that takes many steps and a long time — hours or days of work rather than a single answer — so an AI must stay on track and recover from its own mistakes.
- credit assignment problemcredit assignment problem, n.: The puzzle of figuring out which of many earlier decisions, or which parts of a system, deserve the credit or blame for a final outcome — like working out which moves won a chess game.
- partially observable Markov decision processpartially observable Markov decision process, n.: A version of the standard decision-making model in which the decision-maker cannot see the whole situation — like a poker player who cannot see other hands — and must act on clues and memory.
- multi-agent reinforcement learningmulti-agent reinforcement learning, n.: Reinforcement learning with several AI agents learning at once in a shared world, where each must adapt to the others — as teammates, rivals or both.
- reinforcement learning from human feedbackreinforcement learning from human feedback, n.: A way of training AI using people's judgments: humans compare pairs of answers, a scoring model learns their preferences, and the AI is then trained to produce answers that score well.
Sources
- Zhang, G. et al. (2025). The Landscape of Agentic Reinforcement Learning for LLMs: A Survey. arXiv:2509.02547. arxiv.org/abs/2509.02547
Mentioned in
RL environmentRL environment, n.: A practice world built for training an AI by trial and error: it hands the AI tasks and tools, lets it act, and scores the results so the AI can learn what works., era of experienceera of experience, n.: The prediction, made by David Silver and Richard Sutton in 2025, that AI will soon learn mainly from its own experience of acting in the world rather than from human-written data.