# Multi-hop Knowledge Editing on MQuAKE-3K Source: [Multi-hop Knowledge Editing on MQuAKE-3K]() Collection index: [llms.txt]() Data updated: 2026-03-10T12:15:30.285Z Evaluation of Multi-hop Knowledge Editing performance on the MQuAKE-3K dataset, measured by average performance and edit-wise success rates (1-edit through 4-edit). 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 | Multi-hop Knowledge Editing | | Dataset (including any stated split) | MQuAKE-3K | | Metric | Average Performance | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 22 of 22 | | 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 - [Average Performance](): Higher is better; unit: PERCENT. - [Success Rate (1-edit)](): Higher is better; unit: PERCENT. - [Success Rate (2-edit)](): Higher is better; unit: PERCENT. - [Success Rate (3-edit)](): Higher is better; unit: PERCENT. - [Success Rate (4-edit)](): Higher is better; unit: PERCENT. - [Average Performance (GPT-J)](): Higher is better; unit: PERCENT. - [Average Performance (Qwen3-8B)](): Higher is better; unit: PERCENT. - [Success Rate (1-edit) (GPT-J)](): Higher is better; unit: PERCENT. - [Success Rate (2-edit) (GPT-J)](): Higher is better; unit: PERCENT. - [Success Rate (3-edit) (GPT-J)](): Higher is better; unit: PERCENT. - [Success Rate (4-edit) (GPT-J)](): Higher is better; unit: PERCENT. - [Success Rate (1-edit) (Qwen3-8B)](): Higher is better; unit: PERCENT. - [Success Rate (2-edit) (Qwen3-8B)](): Higher is better; unit: PERCENT. - [Success Rate (3-edit) (Qwen3-8B)](): Higher is better; unit: PERCENT. - [Success Rate (4-edit) (Qwen3-8B)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Average Performance | Result date | Report paper | | --- | --- | --- | --- | | Base | 99.17 | Not recorded | 68fa7c061e209131cea8bf14 | | Base | 98.42 | Not recorded | 68fa7c061e209131cea8bf14 | | ACE (AlphaEdit) | 61.54 | Not recorded | 68fa7c061e209131cea8bf14 | | ACE (PMET) | 58.24 | Not recorded | 68fa7c061e209131cea8bf14 | | ACE (AlphaEdit) | 49.24 | Not recorded | 68fa7c061e209131cea8bf14 | | ACE (PMET) | 46.45 | Not recorded | 68fa7c061e209131cea8bf14 | | AlphaEdit | 45.59 | Not recorded | 68fa7c061e209131cea8bf14 | | AlphaEdit | 39.27 | Not recorded | 68fa7c061e209131cea8bf14 | | MEMIT | 38.58 | Not recorded | 68fa7c061e209131cea8bf14 | | PMET | 37.01 | Not recorded | 68fa7c061e209131cea8bf14 | | ROME | 35.04 | Not recorded | 68fa7c061e209131cea8bf14 | | ROME | 28.79 | Not recorded | 68fa7c061e209131cea8bf14 | | PMET | 20.78 | Not recorded | 68fa7c061e209131cea8bf14 | | MEMIT | 18.67 | Not recorded | 68fa7c061e209131cea8bf14 | | FT | 3.54 | Not recorded | 68fa7c061e209131cea8bf14 | | FT | 2.18 | Not recorded | 68fa7c061e209131cea8bf14 | ### Base Average Performance: 99.17. Result ID: 69b00b516dbb7075b959c9b5. [Reporting paper]() Recorded evaluation configuration: {"Target Model":"Qwen3-8B","Setting":"Few-shot"}. This record does not assert independent reproduction. ### Base Average Performance: 98.42. Result ID: 69b00b516dbb7075b959c9ad. [Reporting paper]() Recorded evaluation configuration: {"Target Model":"GPT-J","Setting":"Few-shot"}. This record does not assert independent reproduction. ### ACE (AlphaEdit) Average Performance: 61.54. Result ID: 69b00b516dbb7075b959c9bc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"Qwen3-8B","Setting":"Few-shot","Backbone Editor":"AlphaEdit"}. This record does not assert independent reproduction. ### ACE (PMET) Average Performance: 58.24. Result ID: 69b00b516dbb7075b959c9bb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"Qwen3-8B","Setting":"Few-shot","Backbone Editor":"PMET"}. This record does not assert independent reproduction. ### ACE (AlphaEdit) Average Performance: 49.24. Result ID: 69b00b516dbb7075b959c9b4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"GPT-J","Setting":"Few-shot","Backbone Editor":"AlphaEdit"}. This record does not assert independent reproduction. ### ACE (PMET) Average Performance: 46.45. Result ID: 69b00b516dbb7075b959c9b3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"GPT-J","Setting":"Few-shot","Backbone Editor":"PMET"}. This record does not assert independent reproduction. ### AlphaEdit Average Performance: 45.59. Result ID: 69b00b516dbb7075b959c9ba. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"Qwen3-8B","Setting":"Few-shot"}. This record does not assert independent reproduction. ### AlphaEdit Average Performance: 39.27. Result ID: 69b00b516dbb7075b959c9b2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"GPT-J","Setting":"Few-shot"}. This record does not assert independent reproduction. ### MEMIT Average Performance: 38.58. Result ID: 69b00b516dbb7075b959c9b0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"GPT-J","Setting":"Few-shot"}. This record does not assert independent reproduction. ### PMET Average Performance: 37.01. Result ID: 69b00b516dbb7075b959c9b1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"GPT-J","Setting":"Few-shot"}. This record does not assert independent reproduction. ### ROME Average Performance: 35.04. Result ID: 69b00b516dbb7075b959c9af. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Target Model":"GPT-J","Setting":"Few-shot"}. This record does not assert independent reproduction. ### ROME Average Performance: 28.79. Result ID: 69b00b516dbb7075b959c9b7. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Target Model":"Qwen3-8B","Setting":"Few-shot"}. This record does not assert independent reproduction. ### PMET Average Performance: 20.78. Result ID: 69b00b516dbb7075b959c9b9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"Qwen3-8B","Setting":"Few-shot"}. This record does not assert independent reproduction. ### MEMIT Average Performance: 18.67. Result ID: 69b00b516dbb7075b959c9b8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Target Model":"Qwen3-8B","Setting":"Few-shot"}. This record does not assert independent reproduction. ### FT Average Performance: 3.54. Result ID: 69b00b516dbb7075b959c9ae. [Reporting paper]() Recorded evaluation configuration: {"Target Model":"GPT-J","Setting":"Few-shot"}. This record does not assert independent reproduction. ### FT Average Performance: 2.18. Result ID: 69b00b516dbb7075b959c9b6. [Reporting paper]() Recorded evaluation configuration: {"Target Model":"Qwen3-8B","Setting":"Few-shot"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.