# Text-to-Image Generation on MS-COCO evaluated after 50-concept erasure

Source: [Text-to-Image Generation on MS-COCO evaluated after 50-concept erasure](<https://www.sota2.com/research/sota/text-to-image-generation-on-ms-coco-evaluated-after-50-concept-erasure>)

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

Data updated: 2026-05-29T08:24:33.121Z

Evaluation of Text-to-Image Generation models using a modified version of the MS-COCO dataset where 50 concepts have been erased from the evaluation set.

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 | Text-to-Image Generation |
| Dataset (including any stated split) | MS-COCO evaluated after 50-concept erasure |
| Metric | CLIP Score |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 13 of 13 |
| 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

- [CLIP Score](<https://www.sota2.com/research/sota/text-to-image-generation-on-ms-coco-evaluated-after-50-concept-erasure.md?metric=CLIP%20Score>): Higher is better; unit: SCALAR.

- [FID](<https://www.sota2.com/research/sota/text-to-image-generation-on-ms-coco-evaluated-after-50-concept-erasure.md?metric=FID>): Lower is better; unit: SCALAR.

## Recorded results

| Method | CLIP Score | Result date | Report paper |
| --- | --- | --- | --- |
| SD v1.4 | 26.56 | Not recorded | 6a191d66dbd7150506f82671 |
| SD v1.4 | 26.53 | Not recorded | 68fa7c231e209131cea8ca66 |
| SPEED | 26.46 | Not recorded | 68fa7c231e209131cea8ca66 |
| Orthogonal Concept Erasure | 26.43 | Not recorded | 6a191d66dbd7150506f82671 |
| FMN | 26.37 | Not recorded | 6a191d66dbd7150506f82671 |
| SPM | 26.32 | Not recorded | 68fa7c231e209131cea8ca66 |
| SPEED | 26.3 | Not recorded | 6a191d66dbd7150506f82671 |
| UCE | 26.19 | Not recorded | 6a191d66dbd7150506f82671 |
| ESD | 25.9 | Not recorded | 6a191d66dbd7150506f82671 |
| CA | 25.66 | Not recorded | 6a191d66dbd7150506f82671 |
| MACE | 25.44 | Not recorded | 6a191d66dbd7150506f82671 |
| SPM w/o FT | 16.22 | Not recorded | 68fa7c231e209131cea8ca66 |
| RECE | 13.8 | Not recorded | 6a191d66dbd7150506f82671 |

### SD v1.4

CLIP Score: 26.56. Result ID: 6a194d31fbc373592c139f70.

[Reporting paper](<https://www.sota2.com/research/paper/6a191d66dbd7150506f82671.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c391e209131cea32e4a.md>)

[Implementation](<https://github.com/CompVis/latent-diffusion>)

Recorded evaluation configuration: {"GPU":"A100"}. This record does not assert independent reproduction.

### SD v1.4

CLIP Score: 26.53. Result ID: 69a7e8626dbb7075b94ca04b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c231e209131cea8ca66.md>)

Recorded evaluation configuration: {"Fine-tuning":"N/A"}. This record does not assert independent reproduction.

### SPEED

CLIP Score: 26.46. Result ID: 69a7e8626dbb7075b94ca04e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c231e209131cea8ca66.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c231e209131cea8ca66.md>)

Recorded evaluation configuration: {"Fine-tuning":"true"}. This record does not assert independent reproduction.

### Orthogonal Concept Erasure

CLIP Score: 26.43. Result ID: 6a194d31fbc373592c139f78.

[Reporting paper](<https://www.sota2.com/research/paper/6a191d66dbd7150506f82671.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a191d66dbd7150506f82671.md>)

Recorded evaluation configuration: {"GPU":"A100"}. This record does not assert independent reproduction.

### FMN

CLIP Score: 26.37. Result ID: 6a194d31fbc373592c139f73.

[Reporting paper](<https://www.sota2.com/research/paper/6a191d66dbd7150506f82671.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ef41e209131cea43246.md>)

Recorded evaluation configuration: {"GPU":"A100"}. This record does not assert independent reproduction.

### SPM

CLIP Score: 26.32. Result ID: 69a7e8626dbb7075b94ca04c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c231e209131cea8ca66.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72761e209131cea56871.md>)

[Implementation](<https://lyumengyao.github.io/projects/spm>)

Recorded evaluation configuration: {"Fine-tuning":"true"}. This record does not assert independent reproduction.

### SPEED

CLIP Score: 26.3. Result ID: 6a194d31fbc373592c139f77.

[Reporting paper](<https://www.sota2.com/research/paper/6a191d66dbd7150506f82671.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c231e209131cea8ca66.md>)

Recorded evaluation configuration: {"GPU":"A100"}. This record does not assert independent reproduction.

### UCE

CLIP Score: 26.19. Result ID: 6a194d31fbc373592c139f74.

[Reporting paper](<https://www.sota2.com/research/paper/6a191d66dbd7150506f82671.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa755b1e209131cea67702.md>)

Recorded evaluation configuration: {"GPU":"A100"}. This record does not assert independent reproduction.

### ESD

CLIP Score: 25.9. Result ID: 6a194d31fbc373592c139f72.

[Reporting paper](<https://www.sota2.com/research/paper/6a191d66dbd7150506f82671.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fc61e209131cea47556.md>)

Recorded evaluation configuration: {"GPU":"A100"}. This record does not assert independent reproduction.

### CA

CLIP Score: 25.66. Result ID: 6a194d31fbc373592c139f71.

[Reporting paper](<https://www.sota2.com/research/paper/6a191d66dbd7150506f82671.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa705c1e209131cea4a98a.md>)

Recorded evaluation configuration: {"GPU":"A100"}. This record does not assert independent reproduction.

### MACE

CLIP Score: 25.44. Result ID: 6a194d31fbc373592c139f75.

[Reporting paper](<https://www.sota2.com/research/paper/6a191d66dbd7150506f82671.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72731e209131cea566f7.md>)

[Implementation](<https://github.com/Shilin-LU/MACE>)

Recorded evaluation configuration: {"GPU":"A100"}. This record does not assert independent reproduction.

### SPM w/o FT

CLIP Score: 16.22. Result ID: 69a7e8626dbb7075b94ca04d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c231e209131cea8ca66.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72761e209131cea56871.md>)

[Implementation](<https://lyumengyao.github.io/projects/spm>)

Recorded evaluation configuration: {"Fine-tuning":"false"}. This record does not assert independent reproduction.

### RECE

CLIP Score: 13.8. Result ID: 6a194d31fbc373592c139f76.

[Reporting paper](<https://www.sota2.com/research/paper/6a191d66dbd7150506f82671.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa756f1e209131cea67e49.md>)

Recorded evaluation configuration: {"GPU":"A100"}. This record does not assert independent reproduction.

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