# Long-context Question Answering on L-Eval QA Source: [Long-context Question Answering on L-Eval QA]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:53:29.002Z Evaluation of model performance on the L-Eval QA dataset for Long-context Question Answering, reporting NQ, NarQA, Crse, and Average 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 | Long-context Question Answering | | Dataset (including any stated split) | L-Eval QA | | Metric | NQ | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 13 of 13 | | 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 - [NQ](): Higher is better; unit: PERCENT. - [NarQA](): Higher is better; unit: PERCENT. - [Crse](): Higher is better; unit: PERCENT. - [Avg Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | NQ | Result date | Report paper | | --- | --- | --- | --- | | DRIFT | 80.73 | Not recorded | 698c0d2fc97a0b3e3b5e127a | | Qwen2.5-Instruct-7B | 80.73 | Not recorded | 698c0d2fc97a0b3e3b5e127a | | Mistral-7B-v0.2 | 79.82 | Not recorded | 698c0d2fc97a0b3e3b5e127a | | LLMLingua-2 | 77.98 | Not recorded | 698c0d2fc97a0b3e3b5e127a | | NaiveRAG | 72.48 | Not recorded | 698c0d2fc97a0b3e3b5e127a | | DRIFT | 71.56 | Not recorded | 698c0d2fc97a0b3e3b5e127a | | DRIFT | 70.64 | Not recorded | 698c0d2fc97a0b3e3b5e127a | | DRIFT | 69.72 | Not recorded | 698c0d2fc97a0b3e3b5e127a | | COCOM | 61.47 | Not recorded | 698c0d2fc97a0b3e3b5e127a | | xRAG | 56.88 | Not recorded | 698c0d2fc97a0b3e3b5e127a | | COCOM | 53.21 | Not recorded | 698c0d2fc97a0b3e3b5e127a | | COCOM | 40.37 | Not recorded | 698c0d2fc97a0b3e3b5e127a | | ICAE | 36.7 | Not recorded | 698c0d2fc97a0b3e3b5e127a | ### DRIFT NQ: 80.73. Result ID: 69963c455698665ec47f408a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Comp. Ratio":"32×","Backbone":"Qwen2.5-Instruct-7B"}. This record does not assert independent reproduction. ### Qwen2.5-Instruct-7B NQ: 80.73. Result ID: 69963c455698665ec47f408c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Comp. Ratio":"1×"}. This record does not assert independent reproduction. ### Mistral-7B-v0.2 NQ: 79.82. Result ID: 69963c455698665ec47f408b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Comp. Ratio":"1×"}. This record does not assert independent reproduction. ### LLMLingua-2 NQ: 77.98. Result ID: 69963c455698665ec47f4080. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Comp. Ratio":"3×","Backbone":"Mistral-7B-v0.2"}. This record does not assert independent reproduction. ### NaiveRAG NQ: 72.48. Result ID: 69963c455698665ec47f4081. [Reporting paper]() Recorded evaluation configuration: {"Comp. Ratio":"-","Backbone":"Mistral-7B-v0.2"}. This record does not assert independent reproduction. ### DRIFT NQ: 71.56. Result ID: 69963c455698665ec47f4089. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Comp. Ratio":"128×","Backbone":"Mistral-7B-v0.2"}. This record does not assert independent reproduction. ### DRIFT NQ: 70.64. Result ID: 69963c455698665ec47f4088. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Comp. Ratio":"64×","Backbone":"Mistral-7B-v0.2"}. This record does not assert independent reproduction. ### DRIFT NQ: 69.72. Result ID: 69963c455698665ec47f4087. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Comp. Ratio":"32×","Backbone":"Mistral-7B-v0.2"}. This record does not assert independent reproduction. ### COCOM NQ: 61.47. Result ID: 69963c455698665ec47f4084. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Comp. Ratio":"4×","Backbone":"Mistral-7B-v0.2"}. This record does not assert independent reproduction. ### xRAG NQ: 56.88. Result ID: 69963c455698665ec47f4082. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Comp. Ratio":"128×","Backbone":"Mistral-7B-v0.2"}. This record does not assert independent reproduction. ### COCOM NQ: 53.21. Result ID: 69963c455698665ec47f4085. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Comp. Ratio":"16×","Backbone":"Mistral-7B-v0.2"}. This record does not assert independent reproduction. ### COCOM NQ: 40.37. Result ID: 69963c455698665ec47f4086. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Comp. Ratio":"128×","Backbone":"Mistral-7B-v0.2"}. This record does not assert independent reproduction. ### ICAE NQ: 36.7. Result ID: 69963c455698665ec47f4083. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Comp. Ratio":"4×","Backbone":"Mistral-7B-v0.2"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.