# Long-context language understanding on LongBench (Mistral-8B-Instruct)

Source: [Long-context language understanding on LongBench (Mistral-8B-Instruct)](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-10T00:18:13.832Z

Evaluation of the Mistral-8B-Instruct model on the LongBench suite, focusing on long-context language understanding capabilities across various sub-tasks.

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 language understanding |
| Dataset (including any stated split) | LongBench Ministral-8B-Instruct |
| Metric | NrtvQA |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 14 of 14 |
| 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

- [NrtvQA](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=NrtvQA>): Higher is better; unit: PERCENT.

- [Qasper](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=Qasper>): Higher is better; unit: PERCENT.

- [MF-en](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=MF-en>): Higher is better; unit: PERCENT.

- [HotpotQA](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=HotpotQA>): Higher is better; unit: PERCENT.

- [2Wiki](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=2Wiki>): Higher is better; unit: PERCENT.

- [Musique](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=Musique>): Higher is better; unit: PERCENT.

- [GovRep](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=GovRep>): Higher is better; unit: PERCENT.

- [QMSum](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=QMSum>): Higher is better; unit: PERCENT.

- [MNews](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=MNews>): Higher is better; unit: PERCENT.

- [TREC](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=TREC>): Higher is better; unit: PERCENT.

- [Trivia](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=Trivia>): Higher is better; unit: PERCENT.

- [SAMSum](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=SAMSum>): Higher is better; unit: PERCENT.

- [PCount](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=PCount>): Higher is better; unit: SCALAR.

- [PRe](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=PRe>): Higher is better; unit: PERCENT.

- [Lcc](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=Lcc>): Higher is better; unit: PERCENT.

- [RB-P](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=RB-P>): Higher is better; unit: PERCENT.

- [Average Score](<https://www.sota2.com/research/sota/long-context-language-understanding-on-longbench-mistral-8b-instruct.md?metric=Average%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | NrtvQA | Result date | Report paper |
| --- | --- | --- | --- |
| StructKV | 30.21 | Not recorded | 69d732b57d0172140ec76cbb |
| StructKV | 30.06 | Not recorded | 69d732b57d0172140ec76cbb |
| GemFilter | 28.06 | Not recorded | 69d732b57d0172140ec76cbb |
| H2O | 27.75 | Not recorded | 69d732b57d0172140ec76cbb |
| H2O | 27.28 | Not recorded | 69d732b57d0172140ec76cbb |
| SnapKV | 26.82 | Not recorded | 69d732b57d0172140ec76cbb |
| SnapKV | 26.72 | Not recorded | 69d732b57d0172140ec76cbb |
| FastKV | 25.9 | Not recorded | 69d732b57d0172140ec76cbb |
| GemFilter | 25.7 | Not recorded | 69d732b57d0172140ec76cbb |
| FastKV | 25.31 | Not recorded | 69d732b57d0172140ec76cbb |
| PyramidInfer | 24.92 | Not recorded | 69d732b57d0172140ec76cbb |
| Full-context | 24.55 | Not recorded | 69d732b57d0172140ec76cbb |
| StreamingLLM | 23.25 | Not recorded | 69d732b57d0172140ec76cbb |
| StreamingLLM | 22.52 | Not recorded | 69d732b57d0172140ec76cbb |

### StructKV

NrtvQA: 30.21. Result ID: 69d8412347f34ce924c044bd.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

Recorded evaluation configuration: {"Prefill":"60%","KV":"20%"}. This record does not assert independent reproduction.

### StructKV

NrtvQA: 30.06. Result ID: 69d8412347f34ce924c044bc.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

Recorded evaluation configuration: {"Prefill":"60%","KV":"10%"}. This record does not assert independent reproduction.

### GemFilter

NrtvQA: 28.06. Result ID: 69d8412347f34ce924c044b9.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74fe1e209131cea651d5.md>)

Recorded evaluation configuration: {"Prefill":"70%","KV":"20%"}. This record does not assert independent reproduction.

### H2O

NrtvQA: 27.75. Result ID: 69d8412347f34ce924c044b3.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa718c1e209131cea51ead.md>)

[Implementation](<https://github.com/FMInference/H2O>)

Recorded evaluation configuration: {"Prefill":"100%","KV":"10%"}. This record does not assert independent reproduction.

### H2O

NrtvQA: 27.28. Result ID: 69d8412347f34ce924c044b4.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa718c1e209131cea51ead.md>)

[Implementation](<https://github.com/FMInference/H2O>)

Recorded evaluation configuration: {"Prefill":"100%","KV":"20%"}. This record does not assert independent reproduction.

### SnapKV

NrtvQA: 26.82. Result ID: 69d8412347f34ce924c044b5.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73b21e209131cea5df89.md>)

[Implementation](<https://github.com/FasterDecoding/SnapKV>)

Recorded evaluation configuration: {"Prefill":"100%","KV":"10%"}. This record does not assert independent reproduction.

### SnapKV

NrtvQA: 26.72. Result ID: 69d8412347f34ce924c044b6.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73b21e209131cea5df89.md>)

[Implementation](<https://github.com/FasterDecoding/SnapKV>)

Recorded evaluation configuration: {"Prefill":"100%","KV":"20%"}. This record does not assert independent reproduction.

### FastKV

NrtvQA: 25.9. Result ID: 69d8412347f34ce924c044bb.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78a41e209131cea79874.md>)

Recorded evaluation configuration: {"Prefill":"60%","KV":"20%"}. This record does not assert independent reproduction.

### GemFilter

NrtvQA: 25.7. Result ID: 69d8412347f34ce924c044b8.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74fe1e209131cea651d5.md>)

Recorded evaluation configuration: {"Prefill":"60%","KV":"10%"}. This record does not assert independent reproduction.

### FastKV

NrtvQA: 25.31. Result ID: 69d8412347f34ce924c044ba.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78a41e209131cea79874.md>)

Recorded evaluation configuration: {"Prefill":"60%","KV":"10%"}. This record does not assert independent reproduction.

### PyramidInfer

NrtvQA: 24.92. Result ID: 69d8412347f34ce924c044b7.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73711e209131cea5c836.md>)

Recorded evaluation configuration: {"Prefill":"60%","KV":"60%"}. This record does not assert independent reproduction.

### Full-context

NrtvQA: 24.55. Result ID: 69d8412347f34ce924c044b0.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

Recorded evaluation configuration: {"Prefill":"100%","KV":"100%"}. This record does not assert independent reproduction.

### StreamingLLM

NrtvQA: 23.25. Result ID: 69d8412347f34ce924c044b2.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72d71e209131cea58d79.md>)

Recorded evaluation configuration: {"Prefill":"100%","KV":"20%"}. This record does not assert independent reproduction.

### StreamingLLM

NrtvQA: 22.52. Result ID: 69d8412347f34ce924c044b1.

[Reporting paper](<https://www.sota2.com/research/paper/69d732b57d0172140ec76cbb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72d71e209131cea58d79.md>)

Recorded evaluation configuration: {"Prefill":"100%","KV":"10%"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
