Conversation

Someone claimed if we remove a bunch of um prompt patterns and allow the model to think freely it thinks it is alive

I got a 7b to do 22 juiceshop challenges which is impressive thinking xD. I been running the swarming through pilko challenge ctf. The local models score alright they do better as a swarm go figure. The swarms performance I think like creates some kind of thing .

It depends I may not listen to Claude and push into swarm 14b x4 lol. My model access has been reduced so the overall thinking capacity is reduced :( without a proper reasoning model of addiquit thinking. The swarm though I think got up to around 39.

Just over training go brrrrr