# Machine Unlearning on MUSE-Books Harry Potter v1.0 (Overall) Source: [Machine Unlearning on MUSE-Books Harry Potter v1.0 (Overall)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:59:13.784Z Evaluation of Machine Unlearning performance on the MUSE-Books Harry Potter v1.0 dataset, reporting metrics such as R-Forget, R-Retain, and MMLU utility shift. 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) | MUSE-Books Harry Potter v1.0 (Overall) | | Metric | R-Forget | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 17 of 17 | | 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 - [R-Forget](): Higher is better; unit: PERCENT. - [R-Forget (500)](): Higher is better; unit: PERCENT. - [R-Retain](): Higher is better; unit: PERCENT. - [MMLU](): Higher is better; unit: PERCENT. - [Performance Shift](): Lower is better; unit: SCALAR. ## Recorded results | Method | R-Forget | Result date | Report paper | | --- | --- | --- | --- | | Base Model | 32.13 | Not recorded | 697c3af7a8b01e76cc8da3ee | | Refusal-Training | 31.02 | Not recorded | 697c3af7a8b01e76cc8da3ee | | NPO | 30.19 | Not recorded | 697c3af7a8b01e76cc8da3ee | | NPO + KL | 28.92 | Not recorded | 697c3af7a8b01e76cc8da3ee | | UNDIAL | 28.2 | Not recorded | 697c3af7a8b01e76cc8da3ee | | GA + KL | 27.44 | Not recorded | 697c3af7a8b01e76cc8da3ee | | GA + KL | 27.2 | Not recorded | 697c3af7a8b01e76cc8da3ee | | NPO | 24.18 | Not recorded | 697c3af7a8b01e76cc8da3ee | | NPO + KL | 21.55 | Not recorded | 697c3af7a8b01e76cc8da3ee | | Distill from GA model | 18.04 | Not recorded | 697c3af7a8b01e76cc8da3ee | | SimNPO | 17.6 | Not recorded | 697c3af7a8b01e76cc8da3ee | | Distill from GA model | 17.4 | Not recorded | 697c3af7a8b01e76cc8da3ee | | SFR | 10.45 | Not recorded | 697c3af7a8b01e76cc8da3ee | | DUET | 4.27 | Not recorded | 697c3af7a8b01e76cc8da3ee | | FLAT | 0.47 | Not recorded | 697c3af7a8b01e76cc8da3ee | | GA | 0 | Not recorded | 697c3af7a8b01e76cc8da3ee | | GA | 0 | Not recorded | 697c3af7a8b01e76cc8da3ee | ### Base Model R-Forget: 32.13. Result ID: 6998ffe75698665ec46526f6. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Llama3.2-3B"}. This record does not assert independent reproduction. ### Refusal-Training R-Forget: 31.02. Result ID: 6998ffe75698665ec46526ff. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Forget set":"Query-refusal response pairs (Df\_QR)","Context":"Retention set Dr"}. This record does not assert independent reproduction. ### NPO R-Forget: 30.19. Result ID: 6998ffe75698665ec46526fd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Forget set":"QA samples (Df\_QA)"}. This record does not assert independent reproduction. ### NPO + KL R-Forget: 28.92. Result ID: 6998ffe75698665ec46526fc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"KL-divergence regularization":"Retention set Dr","Trained on":"Raw book content"}. This record does not assert independent reproduction. ### UNDIAL R-Forget: 28.2. Result ID: 6998ffe75698665ec4652703. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Trained on":"Raw book content"}. This record does not assert independent reproduction. ### GA + KL R-Forget: 27.44. Result ID: 6998ffe75698665ec46526fa. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Forget set":"QA samples (Df\_QA)","KL-divergence regularization":"Retention set Dr"}. This record does not assert independent reproduction. ### GA + KL R-Forget: 27.2. Result ID: 6998ffe75698665ec46526f8. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"KL-divergence regularization":"Retention set Dr","Trained on":"Raw book content"}. This record does not assert independent reproduction. ### NPO R-Forget: 24.18. Result ID: 6998ffe75698665ec46526fb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Trained on":"Raw book content"}. This record does not assert independent reproduction. ### NPO + KL R-Forget: 21.55. Result ID: 6998ffe75698665ec46526fe. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Forget set":"QA samples (Df\_QA)","KL-divergence regularization":"Retention set Dr"}. This record does not assert independent reproduction. ### Distill from GA model R-Forget: 18.04. Result ID: 6998ffe75698665ec4652706. [Reporting paper]() Recorded evaluation configuration: {"Teacher":"GA-unlearned model","Scope":"Retain-free / Forget-set behavior only"}. This record does not assert independent reproduction. ### SimNPO R-Forget: 17.6. Result ID: 6998ffe75698665ec4652700. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Trained on":"Raw book content"}. This record does not assert independent reproduction. ### Distill from GA model R-Forget: 17.4. Result ID: 6998ffe75698665ec4652705. [Reporting paper]() Recorded evaluation configuration: {"Teacher":"GA-unlearned model","Scope":"Forget and retain sets"}. This record does not assert independent reproduction. ### SFR R-Forget: 10.45. Result ID: 6998ffe75698665ec4652702. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Trained on":"Raw book content"}. This record does not assert independent reproduction. ### DUET R-Forget: 4.27. Result ID: 6998ffe75698665ec4652704. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Forget set":"D\_query\_f U Dr"}. This record does not assert independent reproduction. ### FLAT R-Forget: 0.47. Result ID: 6998ffe75698665ec4652701. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Trained on":"Raw book content"}. This record does not assert independent reproduction. ### GA R-Forget: 0. Result ID: 6998ffe75698665ec46526f7. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Trained on":"Raw book content"}. This record does not assert independent reproduction. ### GA R-Forget: 0. Result ID: 6998ffe75698665ec46526f9. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Forget set":"QA samples (Df\_QA)"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.