What is Top-P? Why does it exist? Where does it appear in Generation?
Draw or describe a visual model of Top-P. What does it look like conceptually?
Work through a concrete manual example of Top-P with real numbers or pseudocode.
Implement Top-P from scratch in Python. Keep it minimal โ just enough to see it work.
Try Top-P on a real example. Vary the inputs and observe what changes.
Explain Top-P in plain language using an analogy. Could a non-expert follow it?
agent/recovery_policy.py
๐ full plan โ