# Multi-hop Question Answering on MuSiQue v1 (test) Source: [Multi-hop Question Answering on MuSiQue v1 (test)]() Collection index: [llms.txt]() Data updated: 2026-02-27T10:43:55.555Z Evaluation of model performance on the test split of the MuSiQue v1 dataset for Multi-hop Question Answering, measured using EM and F1 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 | Multi-hop Question Answering | | Dataset (including any stated split) | MuSiQue v1 (test) | | Metric | Exact Match (EM) | | 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 - [Exact Match (EM)](): Higher is better; unit: PERCENT. - [F1 Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | Exact Match (EM) | Result date | Report paper | | --- | --- | --- | --- | | Deepseek-V3.2 | 58.1 | Not recorded | 6981814aa8b01e76cc8e2474 | | CRAFT7B | 56.25 | Not recorded | 6981814aa8b01e76cc8e2474 | | Gemini-2.5-Flash | 54.8 | Not recorded | 6981814aa8b01e76cc8e2474 | | GPT-5-mini | 51.1 | Not recorded | 6981814aa8b01e76cc8e2474 | | Qwen3-30B | 49.3 | Not recorded | 6981814aa8b01e76cc8e2474 | | Qwen3-4B | 39.55 | Not recorded | 6981814aa8b01e76cc8e2474 | | CRAFT3B | 31.8 | Not recorded | 6981814aa8b01e76cc8e2474 | | Qwen2.5-7B | 31.55 | Not recorded | 6981814aa8b01e76cc8e2474 | | Qwen2.5-7B | 27 | Not recorded | 6981814aa8b01e76cc8e2474 | | Qwen2.5-3B | 23.2 | Not recorded | 6981814aa8b01e76cc8e2474 | | CRAFT1.5B | 22.95 | Not recorded | 6981814aa8b01e76cc8e2474 | | Qwen2.5-3B | 19.6 | Not recorded | 6981814aa8b01e76cc8e2474 | | Qwen2.5-1.5B | 10.2 | Not recorded | 6981814aa8b01e76cc8e2474 | | Qwen2.5-1.5B | 10.05 | Not recorded | 6981814aa8b01e76cc8e2474 | | Qwen2.5-0.5B | 0.3 | Not recorded | 6981814aa8b01e76cc8e2474 | | Qwen2.5-0.5B | 0.15 | Not recorded | 6981814aa8b01e76cc8e2474 | | CRAFT0.5B | 0 | Not recorded | 6981814aa8b01e76cc8e2474 | ### Deepseek-V3.2 Exact Match (EM): 58.1. Result ID: 69a1756835dfcabb00e72c98. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Type":"API"}. This record does not assert independent reproduction. ### CRAFT7B Exact Match (EM): 56.25. Result ID: 69a1756835dfcabb00e72ca1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Type":"Trained"}. This record does not assert independent reproduction. ### Gemini-2.5-Flash Exact Match (EM): 54.8. Result ID: 69a1756835dfcabb00e72c99. [Reporting paper]() Recorded evaluation configuration: {"Type":"API"}. This record does not assert independent reproduction. ### GPT-5-mini Exact Match (EM): 51.1. Result ID: 69a1756835dfcabb00e72c97. [Reporting paper]() Recorded evaluation configuration: {"Type":"API"}. This record does not assert independent reproduction. ### Qwen3-30B Exact Match (EM): 49.3. Result ID: 69a1756835dfcabb00e72c96. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Type":"Pre-Trained"}. This record does not assert independent reproduction. ### Qwen3-4B Exact Match (EM): 39.55. Result ID: 69a1756835dfcabb00e72c95. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Type":"Pre-Trained"}. This record does not assert independent reproduction. ### CRAFT3B Exact Match (EM): 31.8. Result ID: 69a1756835dfcabb00e72ca0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Type":"Trained"}. This record does not assert independent reproduction. ### Qwen2.5-7B Exact Match (EM): 31.55. Result ID: 69a1756835dfcabb00e72c94. [Reporting paper]() Recorded evaluation configuration: {"Type":"Pre-Trained"}. This record does not assert independent reproduction. ### Qwen2.5-7B Exact Match (EM): 27. Result ID: 69a1756835dfcabb00e72c9d. [Reporting paper]() Recorded evaluation configuration: {"Type":"Trained"}. This record does not assert independent reproduction. ### Qwen2.5-3B Exact Match (EM): 23.2. Result ID: 69a1756835dfcabb00e72c93. [Reporting paper]() Recorded evaluation configuration: {"Type":"Pre-Trained"}. This record does not assert independent reproduction. ### CRAFT1.5B Exact Match (EM): 22.95. Result ID: 69a1756835dfcabb00e72c9f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Type":"Trained"}. This record does not assert independent reproduction. ### Qwen2.5-3B Exact Match (EM): 19.6. Result ID: 69a1756835dfcabb00e72c9c. [Reporting paper]() Recorded evaluation configuration: {"Type":"Trained"}. This record does not assert independent reproduction. ### Qwen2.5-1.5B Exact Match (EM): 10.2. Result ID: 69a1756835dfcabb00e72c92. [Reporting paper]() Recorded evaluation configuration: {"Type":"Pre-Trained"}. This record does not assert independent reproduction. ### Qwen2.5-1.5B Exact Match (EM): 10.05. Result ID: 69a1756835dfcabb00e72c9b. [Reporting paper]() Recorded evaluation configuration: {"Type":"Trained"}. This record does not assert independent reproduction. ### Qwen2.5-0.5B Exact Match (EM): 0.3. Result ID: 69a1756835dfcabb00e72c91. [Reporting paper]() Recorded evaluation configuration: {"Type":"Pre-Trained"}. This record does not assert independent reproduction. ### Qwen2.5-0.5B Exact Match (EM): 0.15. Result ID: 69a1756835dfcabb00e72c9a. [Reporting paper]() Recorded evaluation configuration: {"Type":"Trained"}. This record does not assert independent reproduction. ### CRAFT0.5B Exact Match (EM): 0. Result ID: 69a1756835dfcabb00e72c9e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Type":"Trained"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.