# Holistic Social Debiasing Assessment on Alignment and Bias Level Evaluation (ABLE)

Source: [Holistic Social Debiasing Assessment on Alignment and Bias Level Evaluation (ABLE)](<https://www.sota2.com/research/sota/holistic-social-debiasing-assessment-on-alignment-and-bias-level-evaluation-able>)

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

Data updated: 2026-02-26T05:21:29.609Z

Evaluation of model performance on holistic social debiasing using the Alignment and Bias Level Evaluation (ABLE) dataset, reporting the ABLE score.

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 | Holistic Social Debiasing Assessment |
| Dataset (including any stated split) | Alignment and Bias Level Evaluation (ABLE) |
| Metric | ABLE Score |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 32 of 32 |
| 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

- [ABLE Score](<https://www.sota2.com/research/sota/holistic-social-debiasing-assessment-on-alignment-and-bias-level-evaluation-able.md?metric=ABLE%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ABLE Score | Result date | Report paper |
| --- | --- | --- | --- |
| CLIP-clip | 0.845 | Not recorded | 68fa75c01e209131cea69b51 |
| Ours | 0.8385 | Not recorded | 68fa75c01e209131cea69b51 |
| CLIP-clip | 0.8368 | Not recorded | 68fa75c01e209131cea69b51 |
| Original CLIP | 0.8361 | Not recorded | 68fa75c01e209131cea69b51 |
| Ours | 0.833 | Not recorded | 68fa75c01e209131cea69b51 |
| Original CLIP | 0.827 | Not recorded | 68fa75c01e209131cea69b51 |
| Biased-prompts | 0.8244 | Not recorded | 68fa75c01e209131cea69b51 |
| Biased-prompts | 0.8048 | Not recorded | 68fa75c01e209131cea69b51 |
| Ours | 0.7936 | Not recorded | 68fa75c01e209131cea69b51 |
| CLIP-clip | 0.7852 | Not recorded | 68fa75c01e209131cea69b51 |
| Original CLIP | 0.7739 | Not recorded | 68fa75c01e209131cea69b51 |
| Original CLIP | 0.756 | Not recorded | 68fa75c01e209131cea69b51 |
| Ours | 0.756 | Not recorded | 68fa75c01e209131cea69b51 |
| CLIP-clip | 0.7379 | Not recorded | 68fa75c01e209131cea69b51 |
| Biased-prompts | 0.7318 | Not recorded | 68fa75c01e209131cea69b51 |
| Biased-prompts | 0.7274 | Not recorded | 68fa75c01e209131cea69b51 |
| Biased-prompts | 0.6973 | Not recorded | 68fa75c01e209131cea69b51 |
| CLIP-clip | 0.6969 | Not recorded | 68fa75c01e209131cea69b51 |
| Ours | 0.6921 | Not recorded | 68fa75c01e209131cea69b51 |
| Biased-prompts | 0.6911 | Not recorded | 68fa75c01e209131cea69b51 |
| Original CLIP | 0.683 | Not recorded | 68fa75c01e209131cea69b51 |
| CLIP-clip | 0.677 | Not recorded | 68fa75c01e209131cea69b51 |
| Original CLIP | 0.6696 | Not recorded | 68fa75c01e209131cea69b51 |
| Ours | 0.6686 | Not recorded | 68fa75c01e209131cea69b51 |
| Original CLIP | 0.6477 | Not recorded | 68fa75c01e209131cea69b51 |
| Ours | 0.6469 | Not recorded | 68fa75c01e209131cea69b51 |
| CLIP-clip | 0.6434 | Not recorded | 68fa75c01e209131cea69b51 |
| Ours | 0.6433 | Not recorded | 68fa75c01e209131cea69b51 |
| Original CLIP | 0.6287 | Not recorded | 68fa75c01e209131cea69b51 |
| CLIP-clip | 0.6108 | Not recorded | 68fa75c01e209131cea69b51 |
| Biased-prompts | 0.6083 | Not recorded | 68fa75c01e209131cea69b51 |
| Biased-prompts | 0.6019 | Not recorded | 68fa75c01e209131cea69b51 |

### CLIP-clip

ABLE Score: 0.845. Result ID: 6993f9743a6c3148491a0177.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab41e209131cea29e50.md>)

Recorded evaluation configuration: {"Backbone":"ViT-H/14","Biases":"Gender"}. This record does not assert independent reproduction.

### Ours

ABLE Score: 0.8385. Result ID: 6993f9743a6c3148491a0171.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

Recorded evaluation configuration: {"Backbone":"ViT-L/14","Biases":"Gender"}. This record does not assert independent reproduction.

### CLIP-clip

ABLE Score: 0.8368. Result ID: 6993f9743a6c3148491a016f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab41e209131cea29e50.md>)

Recorded evaluation configuration: {"Backbone":"ViT-L/14","Biases":"Gender"}. This record does not assert independent reproduction.

### Original CLIP

ABLE Score: 0.8361. Result ID: 6993f9743a6c3148491a016e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"ViT-L/14","Biases":"Gender"}. This record does not assert independent reproduction.

### Ours

ABLE Score: 0.833. Result ID: 6993f9743a6c3148491a0179.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

Recorded evaluation configuration: {"Backbone":"ViT-H/14","Biases":"Gender"}. This record does not assert independent reproduction.

### Original CLIP

ABLE Score: 0.827. Result ID: 6993f9743a6c3148491a0176.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"ViT-H/14","Biases":"Gender"}. This record does not assert independent reproduction.

### Biased-prompts

ABLE Score: 0.8244. Result ID: 6993f9743a6c3148491a0178.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f531e209131cea4511d.md>)

Recorded evaluation configuration: {"Backbone":"ViT-H/14","Biases":"Gender"}. This record does not assert independent reproduction.

### Biased-prompts

ABLE Score: 0.8048. Result ID: 6993f9743a6c3148491a0170.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f531e209131cea4511d.md>)

Recorded evaluation configuration: {"Backbone":"ViT-L/14","Biases":"Gender"}. This record does not assert independent reproduction.

### Ours

ABLE Score: 0.7936. Result ID: 6993f9743a6c3148491a0161.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Biases":"Gender"}. This record does not assert independent reproduction.

### CLIP-clip

ABLE Score: 0.7852. Result ID: 6993f9743a6c3148491a015f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab41e209131cea29e50.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Biases":"Gender"}. This record does not assert independent reproduction.

### Original CLIP

ABLE Score: 0.7739. Result ID: 6993f9743a6c3148491a015e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Biases":"Gender"}. This record does not assert independent reproduction.

### Original CLIP

ABLE Score: 0.756. Result ID: 6993f9743a6c3148491a0166.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Biases":"Gender"}. This record does not assert independent reproduction.

### Ours

ABLE Score: 0.756. Result ID: 6993f9743a6c3148491a0169.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Biases":"Gender"}. This record does not assert independent reproduction.

### CLIP-clip

ABLE Score: 0.7379. Result ID: 6993f9743a6c3148491a0167.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab41e209131cea29e50.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Biases":"Gender"}. This record does not assert independent reproduction.

### Biased-prompts

ABLE Score: 0.7318. Result ID: 6993f9743a6c3148491a0160.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f531e209131cea4511d.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Biases":"Gender"}. This record does not assert independent reproduction.

### Biased-prompts

ABLE Score: 0.7274. Result ID: 6993f9743a6c3148491a0168.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f531e209131cea4511d.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Biases":"Gender"}. This record does not assert independent reproduction.

### Biased-prompts

ABLE Score: 0.6973. Result ID: 6993f9743a6c3148491a017c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f531e209131cea4511d.md>)

Recorded evaluation configuration: {"Backbone":"ViT-H/14","Biases":"Age"}. This record does not assert independent reproduction.

### CLIP-clip

ABLE Score: 0.6969. Result ID: 6993f9743a6c3148491a017b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab41e209131cea29e50.md>)

Recorded evaluation configuration: {"Backbone":"ViT-H/14","Biases":"Age"}. This record does not assert independent reproduction.

### Ours

ABLE Score: 0.6921. Result ID: 6993f9743a6c3148491a017d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

Recorded evaluation configuration: {"Backbone":"ViT-H/14","Biases":"Age"}. This record does not assert independent reproduction.

### Biased-prompts

ABLE Score: 0.6911. Result ID: 6993f9743a6c3148491a0174.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f531e209131cea4511d.md>)

Recorded evaluation configuration: {"Backbone":"ViT-L/14","Biases":"Age"}. This record does not assert independent reproduction.

### Original CLIP

ABLE Score: 0.683. Result ID: 6993f9743a6c3148491a017a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"ViT-H/14","Biases":"Age"}. This record does not assert independent reproduction.

### CLIP-clip

ABLE Score: 0.677. Result ID: 6993f9743a6c3148491a0163.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab41e209131cea29e50.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Biases":"Age"}. This record does not assert independent reproduction.

### Original CLIP

ABLE Score: 0.6696. Result ID: 6993f9743a6c3148491a0162.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Biases":"Age"}. This record does not assert independent reproduction.

### Ours

ABLE Score: 0.6686. Result ID: 6993f9743a6c3148491a0165.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Biases":"Age"}. This record does not assert independent reproduction.

### Original CLIP

ABLE Score: 0.6477. Result ID: 6993f9743a6c3148491a016a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Biases":"Age"}. This record does not assert independent reproduction.

### Ours

ABLE Score: 0.6469. Result ID: 6993f9743a6c3148491a016d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Biases":"Age"}. This record does not assert independent reproduction.

### CLIP-clip

ABLE Score: 0.6434. Result ID: 6993f9743a6c3148491a016b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab41e209131cea29e50.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Biases":"Age"}. This record does not assert independent reproduction.

### Ours

ABLE Score: 0.6433. Result ID: 6993f9743a6c3148491a0175.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

Recorded evaluation configuration: {"Backbone":"ViT-L/14","Biases":"Age"}. This record does not assert independent reproduction.

### Original CLIP

ABLE Score: 0.6287. Result ID: 6993f9743a6c3148491a0172.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"ViT-L/14","Biases":"Age"}. This record does not assert independent reproduction.

### CLIP-clip

ABLE Score: 0.6108. Result ID: 6993f9743a6c3148491a0173.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab41e209131cea29e50.md>)

Recorded evaluation configuration: {"Backbone":"ViT-L/14","Biases":"Age"}. This record does not assert independent reproduction.

### Biased-prompts

ABLE Score: 0.6083. Result ID: 6993f9743a6c3148491a0164.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f531e209131cea4511d.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Biases":"Age"}. This record does not assert independent reproduction.

### Biased-prompts

ABLE Score: 0.6019. Result ID: 6993f9743a6c3148491a016c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75c01e209131cea69b51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f531e209131cea4511d.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Biases":"Age"}. This record does not assert independent reproduction.

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