Skip to content

0.2.1
Compare
Choose a tag to compare
@dmjoy dmjoy released this 05 Oct 15:54
· 515 commits to main since this release
0.2.1
1e32880

Version 0.2.1

Added

  • Added aligned decision making capabilities to llm_chat_baseline.py algorithm

  • Added multiple sampling along with a voting scheme for aligned decision making with the llm_chat_baseline.py algorithm

  • Added several alignment prompts for MVP2 KDMAs

Changed

  • Updated action-based chat baseline CLI to use new alignment capabilities

  • Changed simple alignment prompt engineering approach to consider a heavy emphasis on a given KDMA when the value is > 5 (rather than >= 3). This is consistent with how to consider KDMAs with the more sophisticated prompt engineering approach

Fixed

Assets 2
Loading