News for hackers. Stuff that matters.

A comment on Introducing Agents in Haystack: Make LLMs resolve complex tasks. Parent comment

Re:Introducing Agents in Haystack: Make LLMs resolve complex... by est31 on Monday April 03, 2023 @10:35PM (#35434236)
FTR, people are trying to build systems to compare LLMs with each other based on how well they are at saying "I don't know" (of course knowing is still rewarded higher): https://github.com/manyoso/haltt4llm
[ Parent | on HN ]
Re:Introducing Agents in Haystack: Make LLMs resolve complex... by sr-latch on Monday April 03, 2023 @10:53PM (#35434389)
Would be cool to try to incorporate the previous token's confidence embedding into this process, but that would make training with a triangular attention mask not possible.
[ Parent | on HN ]