3 ms·
i find its the opposite, LLMs can be made to agree with anything.... largely because that agreeability is in their system prompt
by squeefers 8mo ago
i find its the opposite, LLMs can be made to agree with anything.... largely because that agreeability is in their system prompt
- sodapopcan 8mo agoHa, touche!
- batshit_beaver 8mo agoYeah, this. Every conversation inevitably ends with "you're absolutely right!" The number of "you're absolutely right"s per session is roughly how I measure model performance (inverse correlation).