EN
Sign inGet started
Glossary

Reinforcement learning from human feedback (RLHF)

What is Reinforcement learning from human feedback (RLHF)?

RLHF is a training method that makes language models more helpful and safer. People compare and rate several of the model's answers; these judgements are used to train a reward model, which then further trains the language model itself through reinforcement learning. This teaches it, for instance, to answer more clearly or to decline problematic requests. RLHF takes place at the model provider before a model is released – not with individual users.

How KI.ordo helps

The 👍/👎 ratings in KI.ordo do not retrain the model; they show where an agent's knowledge or instructions should be improved.

Learn more →: Reinforcement learning from human feedback (RLHF)

Topics

  • training
  • human feedback
  • language models

Try KI.ordo

Agents, flows and templates in one workspace – free 24-hour demo.

Start demo
← All terms

Related terms

Reinforcement learning (RL)

Reinforcement learning is a branch of machine learning in which an agent learns by trial and error. It acts within an environment, receives rewards for good outcomes and penalties for bad ones, and in this way develops a strategy that maximises its overall reward. Well-known examples include programs that master board games such as Go, robot control and the fine-tuning of language models based on human ratings.

  • learning method
  • reward
  • +1

Instruction Tuning

Instruction tuning is a form of fine-tuning in which a pre-trained language model is trained further on many pairs of instructions and desired responses. This teaches the model to follow tasks phrased in natural language rather than merely continue text. It is a key step in turning a base model into an assistant and is often combined with methods such as RLHF. Example: 'Summarise this text in three sentences' is carried out reliably.

  • Fine-tuning
  • Language models
  • +1

Constitutional AI

Constitutional AI is a training method for language models published by Anthropic in 2022; it has nothing to do with constitutional law. It is based on a written set of principles, the “constitution”: the model critiques and revises its own answers against these principles, then an AI compares pairs of answers, and this feedback serves as the reward signal for reinforcement learning (RLAIF). This reduces the need for human labelling of harmful content; in January 2026 Anthropic published a revised constitution for Claude.

  • Training
  • AI safety
  • +1

Fine-tuning

Fine-tuning means further training an already pre-trained AI model on additional, usually task-specific data. Its internal weights are adjusted so that the model handles, for instance, a particular writing style, specialist vocabulary or output format more reliably. Fine-tuning requires carefully prepared example data, computing power and testing. For many applications, well-written instructions and connected knowledge are sufficient and easier to change.

  • Model training
  • Customisation
  • +1