# Machine Unlearning on TOFU Llama-3.2 Instruct (Avg 1%, 5%, 10% Forget Sets) Source: [Machine Unlearning on TOFU Llama-3.2 Instruct (Avg 1%, 5%, 10% Forget Sets)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:44:59.241Z Evaluation of Machine Unlearning effectiveness on the TOFU Llama-3.2 Instruct configuration, reporting the average performance across the 1%, 5%, and 10% forget sets using metrics like FQ, MU, and ES scores. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Machine Unlearning | | Dataset (including any stated split) | TOFU Llama-3.2 Instruct (average of 1%, 5%, 10% forget sets) | | Metric | FQ Score | | Unit | SCALAR | | Direction | Lower is better | | Results loaded | 18 of 18 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [FQ Score](): Lower is better; unit: SCALAR. - [MU Score](): Higher is better; unit: PERCENT. - [ES (Df)](): Higher is better; unit: PERCENT. - [ES (Dr)](): Higher is better; unit: PERCENT. - [Gib Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | FQ Score | Result date | Report paper | | --- | --- | --- | --- | | GA | -81.257 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | GA | -81.114 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | Full | -13.96 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | GD | -13.959 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | Full | -11.808 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | GD | -8.72 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | TV | -6.174 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | NPO+ | -5.413 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | TV | -5.284 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | NPO | -4.842 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | NPO | -4.508 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | NPO+ | -3.528 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | PerTA-fisher | -1.211 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | PerTA-fisher | -0.867 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | PerTA-grad | -0.686 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | PerTA-grad | -0.669 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | GT | 0 | Not recorded | 697c3b76a8b01e76cc8da7b2 | | GT | 0 | Not recorded | 697c3b76a8b01e76cc8da7b2 | ### GA FQ Score: -81.257. Result ID: 6998f6125698665ec45b1c08. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"3B","Approach":"Training-based"}. This record does not assert independent reproduction. ### GA FQ Score: -81.114. Result ID: 6998f6125698665ec45b1bff. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"1B","Approach":"Training-based"}. This record does not assert independent reproduction. ### Full FQ Score: -13.96. Result ID: 6998f6125698665ec45b1c06. [Reporting paper]() Recorded evaluation configuration: {"Model Size":"3B","Approach":"Reference"}. This record does not assert independent reproduction. ### GD FQ Score: -13.959. Result ID: 6998f6125698665ec45b1c09. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"3B","Approach":"Training-based"}. This record does not assert independent reproduction. ### Full FQ Score: -11.808. Result ID: 6998f6125698665ec45b1bfd. [Reporting paper]() Recorded evaluation configuration: {"Model Size":"1B","Approach":"Reference"}. This record does not assert independent reproduction. ### GD FQ Score: -8.72. Result ID: 6998f6125698665ec45b1c00. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"1B","Approach":"Training-based"}. This record does not assert independent reproduction. ### TV FQ Score: -6.174. Result ID: 6998f6125698665ec45b1c03. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"1B","Approach":"Arithmetic-based"}. This record does not assert independent reproduction. ### NPO+ FQ Score: -5.413. Result ID: 6998f6125698665ec45b1c0b. [Reporting paper]() Recorded evaluation configuration: {"Model Size":"3B","Approach":"Training-based"}. This record does not assert independent reproduction. ### TV FQ Score: -5.284. Result ID: 6998f6125698665ec45b1c0c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"3B","Approach":"Arithmetic-based"}. This record does not assert independent reproduction. ### NPO FQ Score: -4.842. Result ID: 6998f6125698665ec45b1c01. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"1B","Approach":"Training-based"}. This record does not assert independent reproduction. ### NPO FQ Score: -4.508. Result ID: 6998f6125698665ec45b1c0a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"3B","Approach":"Training-based"}. This record does not assert independent reproduction. ### NPO+ FQ Score: -3.528. Result ID: 6998f6125698665ec45b1c02. [Reporting paper]() Recorded evaluation configuration: {"Model Size":"1B","Approach":"Training-based"}. This record does not assert independent reproduction. ### PerTA-fisher FQ Score: -1.211. Result ID: 6998f6125698665ec45b1c0e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"3B","Approach":"Arithmetic-based"}. This record does not assert independent reproduction. ### PerTA-fisher FQ Score: -0.867. Result ID: 6998f6125698665ec45b1c05. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"1B","Approach":"Arithmetic-based"}. This record does not assert independent reproduction. ### PerTA-grad FQ Score: -0.686. Result ID: 6998f6125698665ec45b1c04. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"1B","Approach":"Arithmetic-based"}. This record does not assert independent reproduction. ### PerTA-grad FQ Score: -0.669. Result ID: 6998f6125698665ec45b1c0d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"3B","Approach":"Arithmetic-based"}. This record does not assert independent reproduction. ### GT FQ Score: 0. Result ID: 6998f6125698665ec45b1bfe. [Reporting paper]() Recorded evaluation configuration: {"Model Size":"1B","Approach":"Reference"}. This record does not assert independent reproduction. ### GT FQ Score: 0. Result ID: 6998f6125698665ec45b1c07. [Reporting paper]() Recorded evaluation configuration: {"Model Size":"3B","Approach":"Reference"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.