Hacker News Reader

Roughly analogous process to the temperature parameter in simulated annealing, I think. Low temperature equates to a greedy algorithm (i.e. always take the most likely next word fragment) - higher temps. increase the probability of taking a lower probability fragment.

View on HN

Basically exploration vs exploitation

I’m not sure if comparing it to the greedy algorithm is the correct way to think about it. Even after applying the temperature, if you use greedy decoding, you’ll still pick the output that has the highest probability. Where the concept of temperature is effective though is when you use a sampling method like sampling from a multinomial distribution, nucleus sampling, etc.

The analogy is apt. In SA, the temperature is used to make the distribution sharper or fatter. In this article, the temperature is used to make the distribution sharper or fatter.

Reply on news.ycombinator.com