# End-to-end Question Answering on InfoSeek 25-sample perturbed subset

Source: [End-to-end Question Answering on InfoSeek 25-sample perturbed subset](<https://www.sota2.com/research/sota/end-to-end-question-answering-on-infoseek-25-sample-perturbed-subset>)

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

Data updated: 2026-02-26T06:02:25.261Z

Evaluation of End-to-end Question Answering performance on a specific 25-sample perturbed subset of the InfoSeek dataset, measuring robustness across various transformations like rotation, noise, and blurring.

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 | End-to-end Question Answering |
| Dataset (including any stated split) | InfoSeek 25-sample perturbed subset |
| Metric | Rotation |
| 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

- [Rotation](<https://www.sota2.com/research/sota/end-to-end-question-answering-on-infoseek-25-sample-perturbed-subset.md?metric=Rotation>): Higher is better; unit: PERCENT.

- [Flip](<https://www.sota2.com/research/sota/end-to-end-question-answering-on-infoseek-25-sample-perturbed-subset.md?metric=Flip>): Higher is better; unit: PERCENT.

- [Brightness](<https://www.sota2.com/research/sota/end-to-end-question-answering-on-infoseek-25-sample-perturbed-subset.md?metric=Brightness>): Higher is better; unit: PERCENT.

- [Blur](<https://www.sota2.com/research/sota/end-to-end-question-answering-on-infoseek-25-sample-perturbed-subset.md?metric=Blur>): Higher is better; unit: PERCENT.

- [Gaussian Noise](<https://www.sota2.com/research/sota/end-to-end-question-answering-on-infoseek-25-sample-perturbed-subset.md?metric=Gaussian%20Noise>): Higher is better; unit: PERCENT.

- [Crop](<https://www.sota2.com/research/sota/end-to-end-question-answering-on-infoseek-25-sample-perturbed-subset.md?metric=Crop>): Higher is better; unit: PERCENT.

- [Expand](<https://www.sota2.com/research/sota/end-to-end-question-answering-on-infoseek-25-sample-perturbed-subset.md?metric=Expand>): Higher is better; unit: PERCENT.

- [Overlay](<https://www.sota2.com/research/sota/end-to-end-question-answering-on-infoseek-25-sample-perturbed-subset.md?metric=Overlay>): Higher is better; unit: PERCENT.

- [Watermark](<https://www.sota2.com/research/sota/end-to-end-question-answering-on-infoseek-25-sample-perturbed-subset.md?metric=Watermark>): Higher is better; unit: PERCENT.

- [RealWorld](<https://www.sota2.com/research/sota/end-to-end-question-answering-on-infoseek-25-sample-perturbed-subset.md?metric=RealWorld>): Higher is better; unit: PERCENT.

- [Average Accuracy](<https://www.sota2.com/research/sota/end-to-end-question-answering-on-infoseek-25-sample-perturbed-subset.md?metric=Average%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Rotation | Result date | Report paper |
| --- | --- | --- | --- |
| Oracle | 52 | Not recorded | 6992a4c29ebc499c97406dff |
| V-QPP(SFT) | 48 | Not recorded | 6992a4c29ebc499c97406dff |
| Imperfect | 44 | Not recorded | 6992a4c29ebc499c97406dff |
| V-QPP | 32 | Not recorded | 6992a4c29ebc499c97406dff |

### Oracle

Rotation: 52. Result ID: 699a64ff5698665ec4ed2d58.

[Reporting paper](<https://www.sota2.com/research/paper/6992a4c29ebc499c97406dff.md>)

Recorded evaluation configuration: {"Pre-processing strategy":"Ideal correction","Retrieval API":"Google Lens API"}. This record does not assert independent reproduction.

### V-QPP(SFT)

Rotation: 48. Result ID: 699a64ff5698665ec4ed2d5a.

[Reporting paper](<https://www.sota2.com/research/paper/6992a4c29ebc499c97406dff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6992a4c29ebc499c97406dff.md>)

Recorded evaluation configuration: {"Pre-processing strategy":"Agentic","Agent training protocol":"SFT","Retrieval API":"Google Lens API"}. This record does not assert independent reproduction.

### Imperfect

Rotation: 44. Result ID: 699a64ff5698665ec4ed2d57.

[Reporting paper](<https://www.sota2.com/research/paper/6992a4c29ebc499c97406dff.md>)

Recorded evaluation configuration: {"Pre-processing strategy":"None (no pre-processing)","Retrieval API":"Google Lens API"}. This record does not assert independent reproduction.

### V-QPP

Rotation: 32. Result ID: 699a64ff5698665ec4ed2d59.

[Reporting paper](<https://www.sota2.com/research/paper/6992a4c29ebc499c97406dff.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6992a4c29ebc499c97406dff.md>)

Recorded evaluation configuration: {"Pre-processing strategy":"Agentic","Agent training protocol":"Zero-shot (no training)","Retrieval API":"Google Lens API"}. This record does not assert independent reproduction.

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