Sign inSign in
ConceptAI & ML

Reinforcement learning from human feedback

A technique that trains a reward model directly from human feedback and uses reinforcement learning to optimize an agent's policy.
as Steven Feng defines itStanford UniversityThe wording closest to all 11 definitions
Watch it explained

Know first

Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.

Opens up

Concepts that name this one as a prerequisite.

Lecturers

9

Each in their own terms, from the easiest level.

  • Andrej Karpathy

    Independent · Undergraduate

    RLHF is a training stage that optimizes a language model policy using rewards predicted by a reward model trained on human preferences.
    • 15:15State of GPT: Pretraining, SFT, Reward Modeling & RLHF Pipeline
    • 22:09Introduction to Large Language Models
    • 2:51:15Deep Dive into Large Language Models: Pretraining to RLHF
  • Emma Brunskill

    Stanford University · Undergraduate

    Reinforcement learning performed using a learned reward signal derived from human feedback.
    • 8:51Introduction to Reinforcement Learning (CS234, 2024, Lecture 1)
  • Jason Wei & Hyung Won Chung

    Stanford University · Undergraduate

    An approach that uses a learned reward model as an objective function instead of strict maximum likelihood estimation.
    • 1:14:05Intuitions on Language Models and the History of Transformer Architectures
  • Geoffrey Hinton

    University of Toronto · Undergraduate

    Training a model using human feedback to avoid wrong outputs or reward it for solving puzzles.
    • 1:03:38Digital vs. Biological Intelligence (Vienna Gödel Lecture)
  • Melanie Mitchell

    Santa Fe Institute · Undergraduate

    Training a pre-trained transformer model using human ratings of various generated outputs to guide the model toward preferred responses.
    • 33:57The Future of Artificial Intelligence
  • Chelsea Finn

    Stanford University · Graduate

    A post-training technique using reinforcement learning based on human feedback that can sometimes hurt model calibration.
    • 26:44Frontiers of Deep Reinforcement Learning
  • Eric Mitchell

    UC Berkeley · Graduate

    A training methodology that uses human preference feedback to guide language model behavior.
    • 51:12Reinforcement Learning from Human Feedback: Algorithms & Applications
  • Nathan Lambert

    Stanford University · Graduate

    Specific technical tool for training ML models from human data.
    • 10:38Aligning Open Language Models

Lectures that assume it

1

Connections

19

The lectures that define it or assume it, and the concepts next to it.

9 lecturers. Definitions are written by AI from each lecture’s transcript. A time with a play mark starts the lecture at the moment the lecturer says it.