Sign inSign in
ConceptAI & ML

Direct preference optimization

A method to optimize language models directly on preference data without training a separate reward model.
as Charlie Hou defines itGoogle Research & TechTalksThe wording closest to all 6 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

5

Each in their own terms, from the easiest level.

  • Steven Feng

    Stanford University · Undergraduate

    DPO is an RL-free approach that directly trains the model to prefer human outputs to rank higher without requiring a separate reward model.
    • 41:04Overview of Transformers (Stanford CS25: Transformers United V6)
    • 22:24Overview of Transformers
  • Ming Ding

    Stanford University · Undergraduate

    An alignment method using preference data pairs to update the model without requiring a separate reward model.
    • 27:00From Large Language Models to Large Multimodal Models
  • Eric Mitchell

    UC Berkeley · Graduate

    An algorithm that simplifies RLHF by reparameterizing the reward model directly into the policy, eliminating the need for an explicit reward model training and RL phase.
    • 25:00Reinforcement Learning from Human Feedback: Algorithms & Applications
  • Nathan Lambert

    Stanford University · Graduate

    An algorithm that optimizes language models directly from preferences using gradient ascent without fitting a separate reward model.
    • 43:55Aligning Open Language Models

Lectures that assume it

1

Connections

17

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

5 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.