# Autonomous Driving Question Answering on LingoQA (val)

Source: [Autonomous Driving Question Answering on LingoQA (val)](<https://www.sota2.com/research/sota/autonomous-driving-question-answering-on-lingoqa-val>)

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

Data updated: 2026-03-09T07:09:24.333Z

Evaluation of model performance on the validation split of the LingoQA dataset specifically for Autonomous Driving Question Answering, reporting metrics such as Lingo-J, BLEU, METEOR, and CIDEr.

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 | Autonomous Driving Question Answering |
| Dataset (including any stated split) | LingoQA (val) |
| Metric | Lingo-J |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 6 of 6 |
| 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

- [Lingo-J](<https://www.sota2.com/research/sota/autonomous-driving-question-answering-on-lingoqa-val.md?metric=Lingo-J>): Higher is better; unit: SCALAR.

- [BLEU](<https://www.sota2.com/research/sota/autonomous-driving-question-answering-on-lingoqa-val.md?metric=BLEU>): Higher is better; unit: SCALAR.

- [METEOR](<https://www.sota2.com/research/sota/autonomous-driving-question-answering-on-lingoqa-val.md?metric=METEOR>): Higher is better; unit: SCALAR.

- [CIDEr](<https://www.sota2.com/research/sota/autonomous-driving-question-answering-on-lingoqa-val.md?metric=CIDEr>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Lingo-J | Result date | Report paper |
| --- | --- | --- | --- |
| LEO | 61 | Not recorded | 68fa768c1e209131cea6d9ff |
| LingoQA | 60.8 | Not recorded | 68fa768c1e209131cea6d9ff |
| LingoQA | 59.8 | Not recorded | 68fa768c1e209131cea6d9ff |
| InternVL | 58 | Not recorded | 68fa768c1e209131cea6d9ff |
| BLIP 2 | 52.2 | Not recorded | 68fa768c1e209131cea6d9ff |
| LLaVA-1.5 | 51 | Not recorded | 68fa768c1e209131cea6d9ff |

### LEO

Lingo-J: 61. Result ID: 69ae712a6dbb7075b956076d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa768c1e209131cea6d9ff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa768c1e209131cea6d9ff.md>)

Recorded evaluation configuration: {"N (number of frames used during training)":"2"}. This record does not assert independent reproduction.

### LingoQA

Lingo-J: 60.8. Result ID: 69ae712a6dbb7075b956076c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa768c1e209131cea6d9ff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74fb1e209131cea65110.md>)

Recorded evaluation configuration: {"N (number of frames used during training)":"5"}. This record does not assert independent reproduction.

### LingoQA

Lingo-J: 59.8. Result ID: 69ae712a6dbb7075b956076b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa768c1e209131cea6d9ff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74fb1e209131cea65110.md>)

Recorded evaluation configuration: {"N (number of frames used during training)":"3"}. This record does not assert independent reproduction.

### InternVL

Lingo-J: 58. Result ID: 69ae712a6dbb7075b956076a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa768c1e209131cea6d9ff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71de1e209131cea5370d.md>)

[Implementation](<https://github.com/OpenGVLab/InternVL>)

Recorded evaluation configuration: {"N (number of frames used during training)":"5"}. This record does not assert independent reproduction.

### BLIP 2

Lingo-J: 52.2. Result ID: 69ae712a6dbb7075b9560768.

[Reporting paper](<https://www.sota2.com/research/paper/68fa768c1e209131cea6d9ff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fb71e209131cea46fe6.md>)

[Implementation](<https://github.com/salesforce/LAVIS/tree/main/projects/blip2>)

Recorded evaluation configuration: {"N (number of frames used during training)":"1"}. This record does not assert independent reproduction.

### LLaVA-1.5

Lingo-J: 51. Result ID: 69ae712a6dbb7075b9560769.

[Reporting paper](<https://www.sota2.com/research/paper/68fa768c1e209131cea6d9ff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa732e1e209131cea5af59.md>)

[Implementation](<https://llava-vl.github.io/>)

Recorded evaluation configuration: {"N (number of frames used during training)":"5"}. This record does not assert independent reproduction.

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