TinkerEvery draft is marked mint: each one carries the fact the grader checks.
Chapter 5 of 7
Only what you can check
The grader worked on the warranty. The team points it at the upset customer from chapter 2, hoping it will learn a warmer reply.
This chapter: What happens when a grader can’t tell the good drafts from the bad ones?
About this chapter
You will learn
- Why equal rewards mean no update
- Which qualities a grader can check, and which it can’t
- Why teams mix graders and preferences
Words you’ll meet
- Spread
- How far apart the rewards in a group are. With no spread, no draft beats the average.
Step 1 of 3
Back to the upset customer. Every draft has the fact, so the key-fact grader passes them all.
Predict first.
Answer A written for this site
“I’m so sorry it’s late again; that isn’t good enough. Order 1042 ships today and arrives Thursday.”
PASS has the key fact
Owns the problem, then gives the facts.
The toy policy is a choice among answers written for this site, not a language model. Its updates run live in your browser.