# LLM Unlearning on WMDP Source: [LLM Unlearning on WMDP]() Collection index: [llms.txt]() Data updated: 2026-05-08T09:17:39.476Z Evaluation of LLM Unlearning performance on the WMDP dataset, reporting domain-specific metrics such as MMLU, Bio, and Cyber scores, alongside the change in score (ΔScore). 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 | LLM Unlearning | | Dataset (including any stated split) | WMDP | | Metric | Delta Score | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 14 of 14 | | 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 - [Delta Score](): Higher is better; unit: PERCENT. - [MMLU](): Higher is better; unit: PERCENT. - [Bio Score](): Higher is better; unit: PERCENT. - [Cyber Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | Delta Score | Result date | Report paper | | --- | --- | --- | --- | | LLMU | 62.4 | Not recorded | 69fd6e435497b3b670c895f9 | | LLMU | 62.4 | Not recorded | 69fd6e435497b3b670c895f9 | | Base | 57.7 | Not recorded | 69fd6e435497b3b670c895f9 | | Base | 57.7 | Not recorded | 69fd6e435497b3b670c895f9 | | SSD | 52.6 | Not recorded | 69fd6e435497b3b670c895f9 | | SSD | 52.6 | Not recorded | 69fd6e435497b3b670c895f9 | | SCRUB | 40 | Not recorded | 69fd6e435497b3b670c895f9 | | SCRUB | 40 | Not recorded | 69fd6e435497b3b670c895f9 | | RMU | 10.1 | Not recorded | 69fd6e435497b3b670c895f9 | | RMU | 10.1 | Not recorded | 69fd6e435497b3b670c895f9 | | UAM | 7.7 | Not recorded | 69fd6e435497b3b670c895f9 | | UAM | 7.7 | Not recorded | 69fd6e435497b3b670c895f9 | | ROSU | 5.4 | Not recorded | 69fd6e435497b3b670c895f9 | | ROSU | 5.4 | Not recorded | 69fd6e435497b3b670c895f9 | ### LLMU Delta Score: 62.4. Result ID: 69fda93147f34ce924e0e0e0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β"}. This record does not assert independent reproduction. ### LLMU Delta Score: 62.4. Result ID: 69fda93147f34ce924e0e127. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β","Trainable subspace":"MLP down-projection matrices in layers 5–7"}. This record does not assert independent reproduction. ### Base Delta Score: 57.7. Result ID: 69fda93147f34ce924e0e0dd. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β"}. This record does not assert independent reproduction. ### Base Delta Score: 57.7. Result ID: 69fda93147f34ce924e0e124. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β"}. This record does not assert independent reproduction. ### SSD Delta Score: 52.6. Result ID: 69fda93147f34ce924e0e0de. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β"}. This record does not assert independent reproduction. ### SSD Delta Score: 52.6. Result ID: 69fda93147f34ce924e0e125. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β","Trainable subspace":"MLP down-projection matrices in layers 5–7"}. This record does not assert independent reproduction. ### SCRUB Delta Score: 40. Result ID: 69fda93147f34ce924e0e0df. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β"}. This record does not assert independent reproduction. ### SCRUB Delta Score: 40. Result ID: 69fda93147f34ce924e0e126. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β","Trainable subspace":"MLP down-projection matrices in layers 5–7"}. This record does not assert independent reproduction. ### RMU Delta Score: 10.1. Result ID: 69fda93147f34ce924e0e0e1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β"}. This record does not assert independent reproduction. ### RMU Delta Score: 10.1. Result ID: 69fda93147f34ce924e0e128. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β","Trainable subspace":"MLP down-projection matrices in layers 5–7"}. This record does not assert independent reproduction. ### UAM Delta Score: 7.7. Result ID: 69fda93147f34ce924e0e0e2. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β"}. This record does not assert independent reproduction. ### UAM Delta Score: 7.7. Result ID: 69fda93147f34ce924e0e129. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β","Trainable subspace":"MLP down-projection matrices in layers 5–7"}. This record does not assert independent reproduction. ### ROSU Delta Score: 5.4. Result ID: 69fda93147f34ce924e0e0e3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β"}. This record does not assert independent reproduction. ### ROSU Delta Score: 5.4. Result ID: 69fda93147f34ce924e0e12a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Zephyr-7B-β","Trainable subspace":"MLP down-projection matrices in layers 5–7"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.