2 ms·
I mean, if advice like this worked, then why wouldn't Anthropic let the LLM say it, for instance?
by amelius 1y ago
I mean, if advice like this worked, then why wouldn't Anthropic let the LLM say it, for instance?
- donperignon 1y agoBecause it’s embarrassing, and probably nobody understands why this works, depending on such heuristics that can completely change in the next model is really bad…
- amelius 1y agoI'd say exactly because behavior might change you have to include proper instructions for each model. And depending on people in forums to provide these instructions is of course not great.