This is quite interesting, but as always with LLMs, you have to ask yourself whether the preferences the system appears to reveal are consistent with what the system does in other contexts.
This is quite interesting, but as always with LLMs, you have to ask yourself whether the preferences the system appears to reveal are consistent with what the system does in other contexts.