The temperature dial
The model has read 'The weather today is'. Its scores never change. How often does the reply say something other than sunny - and what decides that?
Five words sit in a raffle drum for the next word after 'The weather today is', and their scores never change. The panels show each word's share of the tickets, the share that misses sunny, and a thousand draws from the drum. Shake it harder and the draws spread out; cool it and they pile onto sunny.
Board: Inference →On a narrow screen the three panels stack below the two dials, and the table scrolls sideways.
try this Drag How hard the drum is shaken from T = 0.1 to 5 and watch Outside the top word climb.
Included with a plan
This lab is included with a plan. The first three lessons of every track are free, with their boards and videos. A plan unlocks every lesson, board, video, lab and paper for one price.
see also Training language models to follow instructions with human feedback in Papers.