# Object Hallucination Evaluation on POPE (Adversarial/Popular/Random Metrics)

Source: [Object Hallucination Evaluation on POPE (Adversarial/Popular/Random Metrics)](<https://www.sota2.com/research/sota/object-hallucination-evaluation-on-pope-adversarial-popular-random-metrics>)

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

Data updated: 2026-06-18T07:12:17.436Z

Evaluation of model object hallucination capabilities on the POPE dataset, reporting performance based on Adversarial, Popular, and Random prompt conditions.

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 | Object Hallucination Evaluation |
| Dataset (including any stated split) | POPE |
| Metric | Adversarial Score |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 3 of 3 |
| 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

- [Adversarial Score](<https://www.sota2.com/research/sota/object-hallucination-evaluation-on-pope-adversarial-popular-random-metrics.md?metric=Adversarial%20Score>): Higher is better; unit: PERCENT.

- [Popular Score](<https://www.sota2.com/research/sota/object-hallucination-evaluation-on-pope-adversarial-popular-random-metrics.md?metric=Popular%20Score>): Higher is better; unit: PERCENT.

- [Random Score](<https://www.sota2.com/research/sota/object-hallucination-evaluation-on-pope-adversarial-popular-random-metrics.md?metric=Random%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Adversarial Score | Result date | Report paper |
| --- | --- | --- | --- |
| KL-KD | 77.87 | Not recorded | 68fa775a1e209131cea72a2d |
| GKL-KD | 77.8 | Not recorded | 68fa775a1e209131cea72a2d |
| Baseline | 75.05 | Not recorded | 68fa775a1e209131cea72a2d |

### KL-KD

Adversarial Score: 77.87. Result ID: 6a339a4bfbc373592c306657.

[Reporting paper](<https://www.sota2.com/research/paper/68fa775a1e209131cea72a2d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa61b81e209131ce9f5b87.md>)

Recorded evaluation configuration: {"CLIP Model":"KL-KD \[27\]"}. This record does not assert independent reproduction.

### GKL-KD

Adversarial Score: 77.8. Result ID: 6a339a4bfbc373592c306658.

[Reporting paper](<https://www.sota2.com/research/paper/68fa775a1e209131cea72a2d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa775a1e209131cea72a2d.md>)

Recorded evaluation configuration: {"CLIP Model":"GKL-KD"}. This record does not assert independent reproduction.

### Baseline

Adversarial Score: 75.05. Result ID: 6a339a4bfbc373592c306656.

[Reporting paper](<https://www.sota2.com/research/paper/68fa775a1e209131cea72a2d.md>)

Recorded evaluation configuration: {"CLIP Model":"Baseline"}. This record does not assert independent reproduction.

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