Skip to content

Clarification on LIBERO evaluation setup in ForeAct #1

@AdvaithRavishankar

Description

@AdvaithRavishankar

Hi, thanks for sharing this work! I had a question about the LIBERO evaluation/training setup and wanted to clarify a few details.

I am trying to understand how the LIBERO experiments were performed, especially in relation to the policy and image generator components.

From the paper/repo, I was unsure about the following:

  1. For the LIBERO results, did you use pi0.5_libero or pi0.5_base as the starting policy checkpoint?

  2. Was the policy further fine-tuned on LIBERO (Spatial, Object, Goal, LIBERO 10) before evaluation, or was it evaluated zero-shot after training on your dataset?

  3. Was imGen further pretrained or fine-tuned on LIBERO data, or was it only trained on the ForeAct dataset?

Thanks in advance!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions