# Visual Attribution-based Hallucination Repair on RePOPE

Source: [Visual Attribution-based Hallucination Repair on RePOPE](<https://www.sota2.com/research/sota/visual-attribution-based-hallucination-repair-on-repope>)

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

Data updated: 2026-07-09T06:50:15.812Z

Evaluation of model performance on repairing visual hallucinations in generated content, specifically focusing on attribution methods, using the RePOPE dataset.

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 | Visual Attribution-based Hallucination Repair |
| Dataset (including any stated split) | RePOPE |
| Metric | Repair Rate |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 8 of 8 |
| 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

- [Repair Rate](<https://www.sota2.com/research/sota/visual-attribution-based-hallucination-repair-on-repope.md?metric=Repair%20Rate>): Higher is better; unit: PERCENT.

- [Accuracy](<https://www.sota2.com/research/sota/visual-attribution-based-hallucination-repair-on-repope.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [Failure Count](<https://www.sota2.com/research/sota/visual-attribution-based-hallucination-repair-on-repope.md?metric=Failure%20Count>): Lower is better; unit: SCALAR.

- [Memory (GB)](<https://www.sota2.com/research/sota/visual-attribution-based-hallucination-repair-on-repope.md?metric=Memory%20(GB)>): Lower is better; unit: SIZE.

- [Inference Time (s)](<https://www.sota2.com/research/sota/visual-attribution-based-hallucination-repair-on-repope.md?metric=Inference%20Time%20(s)>): Lower is better; unit: TIME.

## Recorded results

| Method | Repair Rate | Result date | Report paper |
| --- | --- | --- | --- |
| TRACE Direct | 96 | Not recorded | 6a4f2ae532fb937314875a6c |
| TRACE Direct | 94.44 | Not recorded | 6a4f2ae532fb937314875a6c |
| COPAIR Direct | 84.44 | Not recorded | 6a4f2ae532fb937314875a6c |
| COPAIR Direct | 82.67 | Not recorded | 6a4f2ae532fb937314875a6c |
| IGOS++ | 72.59 | Not recorded | 6a4f2ae532fb937314875a6c |
| IGOS++ | 70.33 | Not recorded | 6a4f2ae532fb937314875a6c |
| LLaVA-CAM | 69.63 | Not recorded | 6a4f2ae532fb937314875a6c |
| LLaVA-CAM | 68.33 | Not recorded | 6a4f2ae532fb937314875a6c |

### TRACE Direct

Repair Rate: 96. Result ID: 6a4f449960b88bb339e44645.

[Reporting paper](<https://www.sota2.com/research/paper/6a4f2ae532fb937314875a6c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a4f2ae532fb937314875a6c.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B","Protocol":"single point"}. This record does not assert independent reproduction.

### TRACE Direct

Repair Rate: 94.44. Result ID: 6a4f449960b88bb339e44641.

[Reporting paper](<https://www.sota2.com/research/paper/6a4f2ae532fb937314875a6c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a4f2ae532fb937314875a6c.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-VL-3B","Protocol":"single point"}. This record does not assert independent reproduction.

### COPAIR Direct

Repair Rate: 84.44. Result ID: 6a4f449960b88bb339e44640.

[Reporting paper](<https://www.sota2.com/research/paper/6a4f2ae532fb937314875a6c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a4f2ae532fb937314875a6c.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-VL-3B","Protocol":"single point"}. This record does not assert independent reproduction.

### COPAIR Direct

Repair Rate: 82.67. Result ID: 6a4f449960b88bb339e44644.

[Reporting paper](<https://www.sota2.com/research/paper/6a4f2ae532fb937314875a6c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a4f2ae532fb937314875a6c.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B","Protocol":"single point"}. This record does not assert independent reproduction.

### IGOS++

Repair Rate: 72.59. Result ID: 6a4f449960b88bb339e4463f.

[Reporting paper](<https://www.sota2.com/research/paper/6a4f2ae532fb937314875a6c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69d21e209131cea2462c.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-VL-3B","Protocol":"full curve"}. This record does not assert independent reproduction.

### IGOS++

Repair Rate: 70.33. Result ID: 6a4f449960b88bb339e44643.

[Reporting paper](<https://www.sota2.com/research/paper/6a4f2ae532fb937314875a6c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69d21e209131cea2462c.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B","Protocol":"full curve"}. This record does not assert independent reproduction.

### LLaVA-CAM

Repair Rate: 69.63. Result ID: 6a4f449960b88bb339e4463e.

[Reporting paper](<https://www.sota2.com/research/paper/6a4f2ae532fb937314875a6c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75401e209131cea66d4e.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-VL-3B","Protocol":"full curve"}. This record does not assert independent reproduction.

### LLaVA-CAM

Repair Rate: 68.33. Result ID: 6a4f449960b88bb339e44642.

[Reporting paper](<https://www.sota2.com/research/paper/6a4f2ae532fb937314875a6c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75401e209131cea66d4e.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B","Protocol":"full curve"}. This record does not assert independent reproduction.

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