Concept
Rubric-and-rule rewards
Rubric-and-rule rewards are a hybrid reward signal for reinforcement learning that combines two scoring mechanisms: deterministic rule-based checks and softer rubric-based judgments, usually from an LLM grader. Think of it like a test suite with both assert statements…
The rest of “Rubric-and-rule rewards” is a premium feature: every concept in the library gets a precise, practitioner-focused write-up like this one, cross-linked straight from the paper summaries that use it.
Log in to unlock→