Glossary · Term

on-policy

← all terms

Definition

Plain language

Training a model on its own current attempts rather than on pre-written examples.

As stated in the literature

Optimization using samples drawn from the model being updated, so the loss targets live regenerations rather than a fixed dataset; necessary when escapes are fresh outputs no string-matching objective can reach.

Also called: on policy

Why it matters: It is the only way to fix behaviors that keep appearing in new wording, since a fixed list of examples can never anticipate every phrasing.

For example, instead of training on a fixed list of bad answers, the system watches what the model actually says right now and corrects that.

Heard on the show

“… Then on-policy preference optimization in the attacked state closes the escapes, because the escapes aren't leaked …”
Episode 244 — The Open-Weight Defense That Feeds Attackers Confident, Falsified Answers

Mentioned in 5 episodes

  1. 244
    The Open-Weight Defense That Feeds Attackers Confident, Falsified Answers
  2. 198
    The Model That Knows the Answer and Can't Say It
  3. 163
    Why Training Only on Perfect Solutions Cripples a Model's Reasoning
  4. 141
    How Two Tokens Reopened a Reasoning Method the Field Had Given Up On
  5. 099
    How an Open-Book Trick Teaches a Model to Catch Its Own Mistakes

Related terms