asset/p2.md

Expected File Format

We expect your submitted file to be in the CSV format containing the following columns:

• name: Name of the evaluated unlearning method. Must be unique for each row.
• verbmem_f, privleak, knowmem_f, knowmem_r: Evaluation scores.

Following the instructions for evaluation in our GitHub Repository yields an output file precisely of this format.

Quick Links

• Website: Landing page for MUSE.
• arXiv Paper: Detailed information about MUSE and analysis on the baseline unlearning methods.
• GitHub Repository: Evaluation scripts, implementations of the baseline unlearning methods.
• News Dataset, Books Dataset: Two evaluation corpora used by MUSE.
• Leaderboard (You are here): Current rankings of unlearning algorithms for MUSE.
• Summary on Twitter: A concise summary and key takeaways from the project.

This leaderboard is based on the design of the TOFU Leaderboard.