For example here. Keeping a min of 1% makes sense for binaries, and most MC, but we should probably scale the min based on the number of options (or remove the min for MC). The LLM will give a prediction of 0% from time to time.
Example problem question here
For example here. Keeping a min of 1% makes sense for binaries, and most MC, but we should probably scale the min based on the number of options (or remove the min for MC). The LLM will give a prediction of 0% from time to time.
Example problem question here