Benchmark

Unlearning Security Audit

A benchmark constructed within the Kaggle Benchmarking Challenge to evaluate large language models' ability to identify and audit fabricated machine unlearning code. It tests models' semantic understanding of the mathematical validity of training objectives.

Timeline (last 90 days)

Oct 11

A developer built a benchmark called Unlearning Security Audit in the Kaggle Benchmarking Challenge to test whether large language models can detect disguised unlearning code.

Unverified50%

All Facts (1)

Unverified

A developer built a benchmark called Unlearning Security Audit in the Kaggle Benchmarking Challenge to test whether large language models can detect disguised unlearning code.

50%

Source Articles

Can LLMs Audit Machine Unlearning? A Benchmark That Exposes Fake Forgetting CodeOct 11