# Linguistic Suitability Evaluation on INFOLOSSQA (test)

Source: [Linguistic Suitability Evaluation on INFOLOSSQA (test)](<https://www.sota2.com/research/sota/linguistic-suitability-evaluation-on-infolossqa-test>)

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

Data updated: 2026-02-26T05:45:59.329Z

Evaluation of model output linguistic suitability on the INFOLOSSQA test set, focusing on metrics related to rationale quality (givenness, missing/unrelated rationale), jargon usage, standalone quality, and hallucination rates.

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 | Linguistic Suitability Evaluation |
| Dataset (including any stated split) | INFOLOSSQA (test) |
| Metric | Givenness Good (%) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 4 of 4 |
| 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

- [Givenness Good (%)](<https://www.sota2.com/research/sota/linguistic-suitability-evaluation-on-infolossqa-test.md?metric=Givenness%20Good%20(%25)>): Higher is better; unit: PERCENT.

- [Rationale Good (%)](<https://www.sota2.com/research/sota/linguistic-suitability-evaluation-on-infolossqa-test.md?metric=Rationale%20Good%20(%25)>): Higher is better; unit: PERCENT.

- [Rationale Unrelated (%)](<https://www.sota2.com/research/sota/linguistic-suitability-evaluation-on-infolossqa-test.md?metric=Rationale%20Unrelated%20(%25)>): Lower is better; unit: PERCENT.

- [Rationale Missing (%)](<https://www.sota2.com/research/sota/linguistic-suitability-evaluation-on-infolossqa-test.md?metric=Rationale%20Missing%20(%25)>): Lower is better; unit: PERCENT.

- [Jargon Level (++) (%)](<https://www.sota2.com/research/sota/linguistic-suitability-evaluation-on-infolossqa-test.md?metric=Jargon%20Level%20(%2B%2B)%20(%25)>): Lower is better; unit: PERCENT.

- [Jargon Level (+) (%)](<https://www.sota2.com/research/sota/linguistic-suitability-evaluation-on-infolossqa-test.md?metric=Jargon%20Level%20(%2B)%20(%25)>): Lower is better; unit: PERCENT.

- [Jargon Level (-) (%)](<https://www.sota2.com/research/sota/linguistic-suitability-evaluation-on-infolossqa-test.md?metric=Jargon%20Level%20(-)%20(%25)>): Lower is better; unit: PERCENT.

- [Jargon Appropriate Level (%)](<https://www.sota2.com/research/sota/linguistic-suitability-evaluation-on-infolossqa-test.md?metric=Jargon%20Appropriate%20Level%20(%25)>): Higher is better; unit: PERCENT.

- [Standalone Quality Good (%)](<https://www.sota2.com/research/sota/linguistic-suitability-evaluation-on-infolossqa-test.md?metric=Standalone%20Quality%20Good%20(%25)>): Higher is better; unit: PERCENT.

- [Hallucination Rate Good (%)](<https://www.sota2.com/research/sota/linguistic-suitability-evaluation-on-infolossqa-test.md?metric=Hallucination%20Rate%20Good%20(%25)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Givenness Good (%) | Result date | Report paper |
| --- | --- | --- | --- |
| GPT-4 | 82 | Not recorded | 68fa73771e209131cea5cb44 |
| Llama2-70B | 75.2 | Not recorded | 68fa73771e209131cea5cb44 |
| Mistral-7B | 74.5 | Not recorded | 68fa73771e209131cea5cb44 |
| NLI Pipeline | 58.3 | Not recorded | 68fa73771e209131cea5cb44 |

### GPT-4

Givenness Good (%): 82. Result ID: 6992e2563a6c314849122921.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73771e209131cea5cb44.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa727a1e209131cea56a1e.md>)

[Implementation](<https://github.com/openai/evals>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Llama2-70B

Givenness Good (%): 75.2. Result ID: 6992e2563a6c314849122920.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73771e209131cea5cb44.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa701f1e209131cea492fe.md>)

[Implementation](<https://github.com/facebookresearch/llama>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Mistral-7B

Givenness Good (%): 74.5. Result ID: 6992e2563a6c31484912291f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73771e209131cea5cb44.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70d61e209131cea4d909.md>)

[Implementation](<https://github.com/mistralai/mistral-src>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### NLI Pipeline

Givenness Good (%): 58.3. Result ID: 6992e2563a6c314849122922.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73771e209131cea5cb44.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73771e209131cea5cb44.md>)

[Implementation](<https://infolossqa.ikim.nrw/>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

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