17 points | by jxmorris12 1 day ago
1 comments
David looks into the LLM finds the thinking layers and cut duplicates then and put them back to back.
This increases the LLM scores with basically no over head.
Very interesting read.
David looks into the LLM finds the thinking layers and cut duplicates then and put them back to back.
This increases the LLM scores with basically no over head.
Very interesting read.