Glossary · Term

decoder

← all terms

Definition

Plain language

The part of a system that turns an internal representation back into something usable, like words, an image, or a robot action.

As stated in the literature

A module that maps a model's latent representation to a target output modality; in Orca, thin per-modality decoders read one shared world-state latent out as text, image, or action.

Also called: decoders

Why it matters: It lets a single shared internal representation power many different outputs, so the hard learning is done once and reused across text, images, and actions.

For example, the same internal understanding of a scene can be sent through one decoder to describe it in words and another to draw it as a picture.

Heard on the show

“A decoder reads that paragraph and outputs: the PIN is 7249.”
Episode 246 — 160 Perfect Refusals, And The Refusals Were The Leak

Mentioned in 13 episodes

  1. 246
    160 Perfect Refusals, And The Refusals Were The Leak
  2. 226
    How a Speed Feature Lets a Stranger Poison Your AI's Answer
  3. 213
    A Model Learned to Control a Robot by Watching Video It Never Acted On
  4. 206
    How Four-Second Clips Become Hours of Playable AI Soccer
  5. 203
    The Thought a Model Doesn't Say — and the Lens That Reads It
  6. 175
    One Crosscoder Feature Flips a Stalling Chatbot Into a Working Agent
  7. 141
    How Two Tokens Reopened a Reasoning Method the Field Had Given Up On
  8. 127
    What Diffusion Language Models Were Missing: A Map, Not an Algorithm
  9. 108
    The Reasoning Cliff: Why Thinking Longer Makes Models Worse at Exact Step-by-Step Tasks
  10. 074
    How a Fifteen-Hundred-Dollar Training Run Matched Llama and Gemma on Reasoning
  11. 070
    When Models Know the Answer But Say the Wrong Thing Anyway
  12. 041
    When the Iteration Teaches the Model to Skip the Iteration
  13. 027
    When AI Agents Build the Serving Stack: A Bet on Bespoke Infrastructure

Related concepts

Related terms