# Concept-preservation human alignment on DreamBench++

Source: [Concept-preservation human alignment on DreamBench++](<https://www.sota2.com/research/sota/concept-preservation-human-alignment-on-dreambench-plus-plus>)

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

Data updated: 2026-04-03T08:56:34.769Z

Evaluation of model alignment capabilities focusing on concept preservation, using the DreamBench++ dataset, measured by metrics such as H-H, G-H, and D-H.

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 | Concept-preservation human alignment |
| Dataset (including any stated split) | DreamBench++ |
| Metric | H-H Score |
| Unit | SCALAR |
| 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

- [H-H Score](<https://www.sota2.com/research/sota/concept-preservation-human-alignment-on-dreambench-plus-plus.md?metric=H-H%20Score>): Higher is better; unit: SCALAR.

- [G-H Score](<https://www.sota2.com/research/sota/concept-preservation-human-alignment-on-dreambench-plus-plus.md?metric=G-H%20Score>): Higher is better; unit: SCALAR.

- [D-H Score](<https://www.sota2.com/research/sota/concept-preservation-human-alignment-on-dreambench-plus-plus.md?metric=D-H%20Score>): Higher is better; unit: SCALAR.

- [C-H Score](<https://www.sota2.com/research/sota/concept-preservation-human-alignment-on-dreambench-plus-plus.md?metric=C-H%20Score>): Higher is better; unit: SCALAR.

- [S-H Score](<https://www.sota2.com/research/sota/concept-preservation-human-alignment-on-dreambench-plus-plus.md?metric=S-H%20Score>): Higher is better; unit: SCALAR.

- [NearID-H Score](<https://www.sota2.com/research/sota/concept-preservation-human-alignment-on-dreambench-plus-plus.md?metric=NearID-H%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | H-H Score | Result date | Report paper |
| --- | --- | --- | --- |
| Emu2 | 0.763 | Not recorded | 69cf49b50636e5bf4afcdb45 |
| DreamBooth | 0.731 | Not recorded | 69cf49b50636e5bf4afcdb45 |
| Textual Inversion | 0.684 | Not recorded | 69cf49b50636e5bf4afcdb45 |
| DreamBooth LoRA | 0.651 | Not recorded | 69cf49b50636e5bf4afcdb45 |
| Fisher-z (r-bar) | 0.648 | Not recorded | 69cf49b50636e5bf4afcdb45 |
| BLIP-Diffusion | 0.63 | Not recorded | 69cf49b50636e5bf4afcdb45 |
| IP-Adapter-Plus ViT-H | 0.53 | Not recorded | 69cf49b50636e5bf4afcdb45 |
| IP-Adapter ViT-G | 0.485 | Not recorded | 69cf49b50636e5bf4afcdb45 |

### Emu2

H-H Score: 0.763. Result ID: 69cf80a447f34ce924b9836d.

[Reporting paper](<https://www.sota2.com/research/paper/69cf49b50636e5bf4afcdb45.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa732a1e209131cea5add9.md>)

[Implementation](<https://github.com/baaivision/Emu>)

Recorded evaluation configuration: {"T2I Model":"SDXL v1.0"}. This record does not assert independent reproduction.

### DreamBooth

H-H Score: 0.731. Result ID: 69cf80a447f34ce924b9836a.

[Reporting paper](<https://www.sota2.com/research/paper/69cf49b50636e5bf4afcdb45.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ec31e209131cea41d81.md>)

[Implementation](<https://dreambooth.github.io/>)

Recorded evaluation configuration: {"T2I Model":"SD v1.5"}. This record does not assert independent reproduction.

### Textual Inversion

H-H Score: 0.684. Result ID: 69cf80a447f34ce924b98369.

[Reporting paper](<https://www.sota2.com/research/paper/69cf49b50636e5bf4afcdb45.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d2d1e209131cea38190.md>)

Recorded evaluation configuration: {"T2I Model":"SD v1.5"}. This record does not assert independent reproduction.

### DreamBooth LoRA

H-H Score: 0.651. Result ID: 69cf80a447f34ce924b9836b.

[Reporting paper](<https://www.sota2.com/research/paper/69cf49b50636e5bf4afcdb45.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b121e209131cea2bf59.md>)

[Implementation](<https://github.com/microsoft/LoRA>)

Recorded evaluation configuration: {"T2I Model":"SDXL v1.0"}. This record does not assert independent reproduction.

### Fisher-z (r-bar)

H-H Score: 0.648. Result ID: 69cf80a447f34ce924b98370.

[Reporting paper](<https://www.sota2.com/research/paper/69cf49b50636e5bf4afcdb45.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69cf49b50636e5bf4afcdb45.md>)

[Implementation](<https://github.com/Gorluxor/NearID>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### BLIP-Diffusion

H-H Score: 0.63. Result ID: 69cf80a447f34ce924b9836c.

[Reporting paper](<https://www.sota2.com/research/paper/69cf49b50636e5bf4afcdb45.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fe21e209131cea47e53.md>)

[Implementation](<https://github.com/salesforce/LAVIS/tree/main/projects/blip-diffusion>)

Recorded evaluation configuration: {"T2I Model":"SD v1.5"}. This record does not assert independent reproduction.

### IP-Adapter-Plus ViT-H

H-H Score: 0.53. Result ID: 69cf80a447f34ce924b9836e.

[Reporting paper](<https://www.sota2.com/research/paper/69cf49b50636e5bf4afcdb45.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70591e209131cea4a798.md>)

[Implementation](<https://ip-adapter.github.io/>)

Recorded evaluation configuration: {"T2I Model":"SDXL v1.0"}. This record does not assert independent reproduction.

### IP-Adapter ViT-G

H-H Score: 0.485. Result ID: 69cf80a447f34ce924b9836f.

[Reporting paper](<https://www.sota2.com/research/paper/69cf49b50636e5bf4afcdb45.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70591e209131cea4a798.md>)

[Implementation](<https://ip-adapter.github.io/>)

Recorded evaluation configuration: {"T2I Model":"SDXL v1.0"}. This record does not assert independent reproduction.

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