# Multi-hop Question Answering on Average (HotpotQA, 2WikiMultiHopQA, MusiQue) (Cost/Efficiency) Source: [Multi-hop Question Answering on Average (HotpotQA, 2WikiMultiHopQA, MusiQue) (Cost/Efficiency)]() Collection index: [llms.txt]() Data updated: 2026-06-23T11:14:03.900Z Evaluation of multi-hop question answering performance averaged across HotpotQA, 2WikiMultiHopQA, and MusiQue, including metrics for accuracy and operational cost/token usage. 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) | HotpotQA, 2WikiMultiHopQA, and MusiQue Average all | | Metric | Cost per Pass | | Unit | SCALAR | | Direction | Lower is better | | Results loaded | 7 of 7 | | 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 - [Cost per Pass](): Lower is better; unit: SCALAR. - [Accuracy](): Higher is better; unit: PERCENT. - [Cost (cents)](): Lower is better; unit: SCALAR. - [Token Count](): Lower is better; unit: SIZE. ## Recorded results | Method | Cost per Pass | Result date | Report paper | | --- | --- | --- | --- | | Standard RAG | 0.14 | Not recorded | 6a3a1362206542251d7c7406 | | IRCoT | 0.17 | Not recorded | 6a3a1362206542251d7c7406 | | Search-o1 | 0.49 | Not recorded | 6a3a1362206542251d7c7406 | | GDP-RAG | 0.51 | Not recorded | 6a3a1362206542251d7c7406 | | PAR-RAG | 0.65 | Not recorded | 6a3a1362206542251d7c7406 | | Godbole et al. | 1.12 | Not recorded | 6a3a1362206542251d7c7406 | | KnowTrace | 1.57 | Not recorded | 6a3a1362206542251d7c7406 | ### Standard RAG Cost per Pass: 0.14. Result ID: 6a3a6a67fbc373592c3796a8. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### IRCoT Cost per Pass: 0.17. Result ID: 6a3a6a67fbc373592c3796a9. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Search-o1 Cost per Pass: 0.49. Result ID: 6a3a6a67fbc373592c3796ab. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### GDP-RAG Cost per Pass: 0.51. Result ID: 6a3a6a67fbc373592c3796ae. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### PAR-RAG Cost per Pass: 0.65. Result ID: 6a3a6a67fbc373592c3796ac. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Godbole et al. Cost per Pass: 1.12. Result ID: 6a3a6a67fbc373592c3796ad. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"adaptation":"QA first planning stage only"}. This record does not assert independent reproduction. ### KnowTrace Cost per Pass: 1.57. Result ID: 6a3a6a67fbc373592c3796aa. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"mode":"inference-time KG-tracing","self-bootstrapping fine-tuning":"False"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.