Show HN: Predict LLM Predictionsthink-in-sync.com 3 pointsflowingfocus1 year ago2 commentsSaveHideCopy link On HNComments−ggm1yAdversarial play might be interesting. I'm assuming the goal here is to use the feedback as training input, fitness to expectations and like.So a random chance of a bad actor adversary? Or high scores for lying? "The traitor" styleTeam models? I suppose I'm arguing for gamification with a leader board and hubristic bragging rights.−flowingfocusOP1yAdversarial play sounds fun, I might try that out!The LLM we use is mistral-large-latest, we didn't do any training on the data yet
Comments
Adversarial play might be interesting. I'm assuming the goal here is to use the feedback as training input, fitness to expectations and like.
So a random chance of a bad actor adversary? Or high scores for lying? "The traitor" style
Team models? I suppose I'm arguing for gamification with a leader board and hubristic bragging rights.
Adversarial play sounds fun, I might try that out!
The LLM we use is mistral-large-latest, we didn't do any training on the data yet