Edit ‘autogollark’

This commit is contained in:
osmarks
2025-04-19 13:30:04 +00:00
committed by wikimind
parent 4d65097b05
commit 3f5debf412

View File

@@ -23,6 +23,8 @@ Autogollark currently comprises the dataset, the search API server and the [[htt
* Tool capabilities (how to get the data? Examples in context only?!).
* {Local finetune only? Would be more tonally consistent but dumber, I think.
* Temporary bursts of hypercompetence enabled by powerful base model are a key feature. Small model is really repetitive.
* Can additionally finetune on "interesting" blog posts etc (ref https://x.com/QiaochuYuan/status/1913382597381767471).
* Decision theory training data (synthetic, probably) (ref https://arxiv.org/abs/2411.10588).
}
* MCTS over conversations with non-gollark simulacra?