Path: How post-training works
What an RL environment is
See how a task becomes something a model can practise against by itself, and what makes a task unsuitable.
- Starting level
- Some basics
- Time
- 18 minutes
- Public lesson guide
- EN + हिंदी
You will learn to
Say whether a task can be turned into practice a model runs by itself, and why.
Why this matters
Post-training scales only where success can be judged without a person reading every attempt.
Three things to remember
- A task is trainable when a rule can score it without a person.
- Write the success check before the task is called practice.
- Name what the automatic check cannot see.
See one example
Task
Three candidate tasks: a date conversion, a condolence message, a polite greeting.
Try the judgement yourself
Both answers respond to the task above. One of them does the job better. Which, and why?
Make something yourself
Choose one task from your own work or study. Write the input, the exact rule that would decide success without a person, and one quality that rule cannot see.
A useful starting pointFour parts: the task, its input, the automatic check, and the blind spot.
Read further
The source is optional. The complete teaching and practice are available inside AI2Bharat.
Read the primary or official source: OpenAI Gym