ORBiS Tree
KO

ORBiS TREE · ENGLISH DOCUMENT

Concepts →

RLHF

Reinforcement Learning from Human Feedback, a method for aligning model behavior with human preferences.

한국어 English
#Alignment#LLM#RLHF#Reinforcement Learning
Korean source Edit source
English updated 8월 29, 2026 Source article updated 8월 29, 2026 2 sources
This English page is a curated translation layer linked to the Korean source article. Community changes are currently made on the Korean source, where the full revision history and anonymous edit trail are preserved.
RLHF is a family of training methods that uses human judgments to shape model behavior. It became widely associated with instruction-following language models and conversational systems.

How it works

A common pipeline first collects preference comparisons between model outputs, trains a reward or preference model, and then optimizes the language model to produce outputs that score better under that learned signal. Implementations vary and newer preference-optimization methods may avoid some traditional RL steps.

Why it matters

RLHF can improve helpfulness and instruction following, but human feedback is not a perfect measure of truth or safety. Dataset design, annotator incentives, and optimization choices can introduce their own biases.

Related concepts

See also Fine-tuning, Safety Evaluation, and Red Teaming.

SOURCES

Sources

  1. Deep reinforcement learning from human preferencesDeepMind / OpenAI / arXiv
    Open source ↗
  2. Training language models to follow instructions with human feedbackOpenAI / arXiv
    Open source ↗

KNOWLEDGE LINKS

Fine-tuning →

KEEP EXPLORING

Related articles

Latency and ThroughputTwo core serving metrics describing response delay and the amount of work a system completes over time.GPUA highly parallel processor widely used to train and run modern machine-learning models.Context WindowThe amount of tokenized input and generated text a model can consider within one inference context.KV CacheA cache of attention keys and values used to avoid recomputing earlier context during autoregressive generation.InferenceThe process of running a trained model on new input to produce predictions, embeddings, or generated output.Knowledge DistillationA training method in which a smaller student model learns from the outputs or representations of a larger teacher model.
ORBiS Tree English · source-linked translation
Korean source & revision history · Report an issue · Editorial policy