Skip to content

Multi-Step RL with CUBE#139

Open
Masseeh wants to merge 1 commit intoServiceNow:multi-stepfrom
Masseeh:multi-step-cube
Open

Multi-Step RL with CUBE#139
Masseeh wants to merge 1 commit intoServiceNow:multi-stepfrom
Masseeh:multi-step-cube

Conversation

@Masseeh
Copy link
Copy Markdown

@Masseeh Masseeh commented May 1, 2026

CUBE compatible Multi-step RL in PipelineRL.

  • multi-turn rollouts with a single/dense scalar reward per rollout.
  • ray compatible actor.
  • tested on a simple multi-step math cube + cube-harness agents

Cube-harness compatible repo: https://github.com/Masseeh/cube-harness/tree/piplinerl-cube
Cube-standard compatible repo: https://github.com/Masseeh/cube-standard/tree/piplinerl-cube

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant