# Long-context understanding on LongBench E (Functional Metrics & Latency) Source: [Long-context understanding on LongBench E (Functional Metrics & Latency)]() Collection index: [llms.txt]() Data updated: 2026-02-26T06:04:23.584Z Evaluation of language models on the LongBench E suite, reporting performance aggregated across key functional areas (Single Doc QA, Multi Doc QA, Summarization, Code) and including system latency measurements. 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 understanding | | Dataset (including any stated split) | LongBench E | | Metric | Single Doc QA | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 10 of 10 | | 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 - [Single Doc QA](): Higher is better; unit: PERCENT. - [Multi Doc QA](): Higher is better; unit: PERCENT. - [Summarization](): Higher is better; unit: PERCENT. - [Few Shot Score](): Higher is better; unit: PERCENT. - [Syntax Score](): Higher is better; unit: PERCENT. - [Code](): Higher is better; unit: PERCENT. - [Average Score](): Higher is better; unit: PERCENT. - [Latency (ms)](): Lower is better; unit: TIME. ## Recorded results | Method | Single Doc QA | Result date | Report paper | | --- | --- | --- | --- | | TCA-Attention | 52.28 | Not recorded | 693a50246cfd488761a54ba7 | | MInference | 51.7 | Not recorded | 693a50246cfd488761a54ba7 | | LLaMA3.1-8B-Instruct | 51.63 | Not recorded | 693a50246cfd488761a54ba7 | | FlexPrefill | 50.35 | Not recorded | 693a50246cfd488761a54ba7 | | XAttention | 49.96 | Not recorded | 693a50246cfd488761a54ba7 | | FlexPrefill | 49.08 | Not recorded | 693a50246cfd488761a54ba7 | | MInference | 48.8 | Not recorded | 693a50246cfd488761a54ba7 | | Qwen2.5-7B-Instruct | 48.75 | Not recorded | 693a50246cfd488761a54ba7 | | XAttention | 48.5 | Not recorded | 693a50246cfd488761a54ba7 | | TCA-Attention | 48.5 | Not recorded | 693a50246cfd488761a54ba7 | ### TCA-Attention Single Doc QA: 52.28. Result ID: 6996c5ae5698665ec4112a29. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLaMA3.1-8B-Instruct","Context Length":"64K"}. This record does not assert independent reproduction. ### MInference Single Doc QA: 51.7. Result ID: 6996c5ae5698665ec4112a26. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model":"LLaMA3.1-8B-Instruct","Context Length":"64K"}. This record does not assert independent reproduction. ### LLaMA3.1-8B-Instruct Single Doc QA: 51.63. Result ID: 6996c5ae5698665ec4112a25. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model":"LLaMA3.1-8B-Instruct","Attention":"Full","Context Length":"64K"}. This record does not assert independent reproduction. ### FlexPrefill Single Doc QA: 50.35. Result ID: 6996c5ae5698665ec4112a27. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLaMA3.1-8B-Instruct","Context Length":"64K"}. This record does not assert independent reproduction. ### XAttention Single Doc QA: 49.96. Result ID: 6996c5ae5698665ec4112a28. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLaMA3.1-8B-Instruct","Context Length":"64K"}. This record does not assert independent reproduction. ### FlexPrefill Single Doc QA: 49.08. Result ID: 6996c5ae5698665ec4112a2c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-7B-Instruct","Context Length":"64K"}. This record does not assert independent reproduction. ### MInference Single Doc QA: 48.8. Result ID: 6996c5ae5698665ec4112a2b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model":"Qwen2.5-7B-Instruct","Context Length":"64K"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Instruct Single Doc QA: 48.75. Result ID: 6996c5ae5698665ec4112a2a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-7B-Instruct","Attention":"Full","Context Length":"64K"}. This record does not assert independent reproduction. ### XAttention Single Doc QA: 48.5. Result ID: 6996c5ae5698665ec4112a2d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-7B-Instruct","Context Length":"64K"}. This record does not assert independent reproduction. ### TCA-Attention Single Doc QA: 48.5. Result ID: 6996c5ae5698665ec4112a2e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-7B-Instruct","Context Length":"64K"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.