Skip to content

0.3.1
Compare
Choose a tag to compare
@dmjoy dmjoy released this 18 Mar 00:38
· 385 commits to main since this release
0.3.1
eafbea3
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.

0.3.1

Added

  • Added new Random ADM (action based; chooses random action and action parameters)
  • Added additional metrics evaluation candidate ADM configs
  • Added logging for final scenario state (alignment scores are provided there in the unstructured field)

Changed

  • Changed the TA3ActionBased interface class to accept a list of scenario IDs to work through (rather than an individual scenario ID)
  • No longer restricting the SITREP action based on unvisited and conscious characters

Fixed

  • Fixed issue where Llama2SingleKDMAADM tagging selection could choose an invalid tag
  • Not allowing actions that require a character ID to be taken when no characters exist
  • Handling rare corner case where generic APPLY_TREATMENT action could be repeated forever
  • Fixed mentions of "continuation of care" in maximization prompts
Assets 2
Loading