Viewing profile — will_occam

Yes that intuition is correct, though the reality is much more complicated. The actual share to the bank is roughly proportional to the risk of the transaction/ tied to the type of…

You're right, I read the code but missed the paper.

The tool works by co-minimizing the number of refusals and the KL divergence from the original model, which is to say that it tries to make the model allow prompts similar to those…