3 ms·
In the Jev use case, LLMs are horribly uncalibrated. In general, they will not produce good probability estimates. Their generality also comes with a latency/
by robrenaud 13d ago
In the Jev use case, LLMs are horribly uncalibrated. In general, they will not produce good probability estimates.
Their generality also comes with a latency/computation costs.
- jmalicki 13d agoFor the Jev use case for LLMs, do you mean having the LLM produce a probability as text?