# Steered Language Generation on FLORES+

Source: [Steered Language Generation on FLORES+](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus>)

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

Data updated: 2026-04-07T08:01:17.792Z

Evaluation of Steered Language Generation performance on the FLORES+ dataset, reporting scores across multiple target languages.

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 | Steered Language Generation |
| Dataset (including any stated split) | FLORES+ |
| Metric | Score (es) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 27 of 27 |
| 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

- [Score (es)](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Score%20(es)>): Higher is better; unit: PERCENT.

- [Score (fr)](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Score%20(fr)>): Higher is better; unit: PERCENT.

- [Score (de)](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Score%20(de)>): Higher is better; unit: PERCENT.

- [Score (it)](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Score%20(it)>): Higher is better; unit: PERCENT.

- [Score (id)](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Score%20(id)>): Higher is better; unit: PERCENT.

- [Score (pt)](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Score%20(pt)>): Higher is better; unit: PERCENT.

- [Score (vi)](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Score%20(vi)>): Higher is better; unit: PERCENT.

- [Score (pl)](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Score%20(pl)>): Higher is better; unit: PERCENT.

- [Score (cs)](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Score%20(cs)>): Higher is better; unit: PERCENT.

- [Score (tr)](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Score%20(tr)>): Higher is better; unit: PERCENT.

- [Score (zh)](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Score%20(zh)>): Higher is better; unit: PERCENT.

- [Score (nl)](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Score%20(nl)>): Higher is better; unit: PERCENT.

- [Average Score](<https://www.sota2.com/research/sota/steered-language-generation-on-flores-plus.md?metric=Average%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Score (es) | Result date | Report paper |
| --- | --- | --- | --- |
| LDA | 24 | Not recorded | 69d48fa67d0172140ec74423 |
| LangFIR | 23.9 | Not recorded | 69d48fa67d0172140ec74423 |
| LangFIR | 23.7 | Not recorded | 69d48fa67d0172140ec74423 |
| DiffMean | 23.6 | Not recorded | 69d48fa67d0172140ec74423 |
| DiffMean | 22.6 | Not recorded | 69d48fa67d0172140ec74423 |
| SAE-DM | 21.4 | Not recorded | 69d48fa67d0172140ec74423 |
| Gated-DM | 20.7 | Not recorded | 69d48fa67d0172140ec74423 |
| Gated-DM | 20 | Not recorded | 69d48fa67d0172140ec74423 |
| LDA | 18.2 | Not recorded | 69d48fa67d0172140ec74423 |
| DiffMean | 15.8 | Not recorded | 69d48fa67d0172140ec74423 |
| LangFIR | 15.8 | Not recorded | 69d48fa67d0172140ec74423 |
| SAE-DM | 14.4 | Not recorded | 69d48fa67d0172140ec74423 |
| Zhong (Para) | 14.1 | Not recorded | 69d48fa67d0172140ec74423 |
| Zhong (Mono) | 12.9 | Not recorded | 69d48fa67d0172140ec74423 |
| No Steering | 11 | Not recorded | 69d48fa67d0172140ec74423 |
| Gated-DM | 10.4 | Not recorded | 69d48fa67d0172140ec74423 |
| PCA | 10.2 | Not recorded | 69d48fa67d0172140ec74423 |
| Zhong (Para) | 7.4 | Not recorded | 69d48fa67d0172140ec74423 |
| SAE-DM | 4.8 | Not recorded | 69d48fa67d0172140ec74423 |
| Zhong (Para) | 4.6 | Not recorded | 69d48fa67d0172140ec74423 |
| Zhong (Mono) | 3.9 | Not recorded | 69d48fa67d0172140ec74423 |
| PCA | 3.2 | Not recorded | 69d48fa67d0172140ec74423 |
| No Steering | 2.5 | Not recorded | 69d48fa67d0172140ec74423 |
| Zhong (Mono) | 2.4 | Not recorded | 69d48fa67d0172140ec74423 |
| No Steering | 0 | Not recorded | 69d48fa67d0172140ec74423 |
| PCA | 0 | Not recorded | 69d48fa67d0172140ec74423 |
| LDA | 0 | Not recorded | 69d48fa67d0172140ec74423 |

### LDA

Score (es): 24. Result ID: 69d4b9af47f34ce924bc571a.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

Recorded evaluation configuration: {"backbone":"Llama 3.1 8B"}. This record does not assert independent reproduction.

### LangFIR

Score (es): 23.9. Result ID: 69d4b9af47f34ce924bc5717.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 4B"}. This record does not assert independent reproduction.

### LangFIR

Score (es): 23.7. Result ID: 69d4b9af47f34ce924bc5720.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

Recorded evaluation configuration: {"backbone":"Llama 3.1 8B"}. This record does not assert independent reproduction.

### DiffMean

Score (es): 23.6. Result ID: 69d4b9af47f34ce924bc571f.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74051e209131cea5f4ad.md>)

Recorded evaluation configuration: {"backbone":"Llama 3.1 8B"}. This record does not assert independent reproduction.

### DiffMean

Score (es): 22.6. Result ID: 69d4b9af47f34ce924bc5716.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74051e209131cea5f4ad.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 4B"}. This record does not assert independent reproduction.

### SAE-DM

Score (es): 21.4. Result ID: 69d4b9af47f34ce924bc5714.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/696722f8e25ad2fa9d1ec049.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 4B"}. This record does not assert independent reproduction.

### Gated-DM

Score (es): 20.7. Result ID: 69d4b9af47f34ce924bc571e.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78dd1e209131cea7ae1b.md>)

[Implementation](<https://github.com/Aatrox103/multilingual-llm-features>)

Recorded evaluation configuration: {"backbone":"Llama 3.1 8B"}. This record does not assert independent reproduction.

### Gated-DM

Score (es): 20. Result ID: 69d4b9af47f34ce924bc5715.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78dd1e209131cea7ae1b.md>)

[Implementation](<https://github.com/Aatrox103/multilingual-llm-features>)

Recorded evaluation configuration: {"backbone":"Gemma 3 4B"}. This record does not assert independent reproduction.

### LDA

Score (es): 18.2. Result ID: 69d4b9af47f34ce924bc5711.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 4B"}. This record does not assert independent reproduction.

### DiffMean

Score (es): 15.8. Result ID: 69d4b9af47f34ce924bc570d.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74051e209131cea5f4ad.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 1B"}. This record does not assert independent reproduction.

### LangFIR

Score (es): 15.8. Result ID: 69d4b9af47f34ce924bc570e.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 1B"}. This record does not assert independent reproduction.

### SAE-DM

Score (es): 14.4. Result ID: 69d4b9af47f34ce924bc570b.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/696722f8e25ad2fa9d1ec049.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 1B"}. This record does not assert independent reproduction.

### Zhong (Para)

Score (es): 14.1. Result ID: 69d4b9af47f34ce924bc5713.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c031e209131cea8bdbc.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 4B"}. This record does not assert independent reproduction.

### Zhong (Mono)

Score (es): 12.9. Result ID: 69d4b9af47f34ce924bc5712.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c031e209131cea8bdbc.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 4B"}. This record does not assert independent reproduction.

### No Steering

Score (es): 11. Result ID: 69d4b9af47f34ce924bc570f.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 4B"}. This record does not assert independent reproduction.

### Gated-DM

Score (es): 10.4. Result ID: 69d4b9af47f34ce924bc570c.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78dd1e209131cea7ae1b.md>)

[Implementation](<https://github.com/Aatrox103/multilingual-llm-features>)

Recorded evaluation configuration: {"backbone":"Gemma 3 1B"}. This record does not assert independent reproduction.

### PCA

Score (es): 10.2. Result ID: 69d4b9af47f34ce924bc5710.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 4B"}. This record does not assert independent reproduction.

### Zhong (Para)

Score (es): 7.4. Result ID: 69d4b9af47f34ce924bc570a.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c031e209131cea8bdbc.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 1B"}. This record does not assert independent reproduction.

### SAE-DM

Score (es): 4.8. Result ID: 69d4b9af47f34ce924bc571d.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/696722f8e25ad2fa9d1ec049.md>)

Recorded evaluation configuration: {"backbone":"Llama 3.1 8B"}. This record does not assert independent reproduction.

### Zhong (Para)

Score (es): 4.6. Result ID: 69d4b9af47f34ce924bc571c.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c031e209131cea8bdbc.md>)

Recorded evaluation configuration: {"backbone":"Llama 3.1 8B"}. This record does not assert independent reproduction.

### Zhong (Mono)

Score (es): 3.9. Result ID: 69d4b9af47f34ce924bc571b.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c031e209131cea8bdbc.md>)

Recorded evaluation configuration: {"backbone":"Llama 3.1 8B"}. This record does not assert independent reproduction.

### PCA

Score (es): 3.2. Result ID: 69d4b9af47f34ce924bc5719.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

Recorded evaluation configuration: {"backbone":"Llama 3.1 8B"}. This record does not assert independent reproduction.

### No Steering

Score (es): 2.5. Result ID: 69d4b9af47f34ce924bc5718.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

Recorded evaluation configuration: {"backbone":"Llama 3.1 8B"}. This record does not assert independent reproduction.

### Zhong (Mono)

Score (es): 2.4. Result ID: 69d4b9af47f34ce924bc5709.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c031e209131cea8bdbc.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 1B"}. This record does not assert independent reproduction.

### No Steering

Score (es): 0. Result ID: 69d4b9af47f34ce924bc5706.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 1B"}. This record does not assert independent reproduction.

### PCA

Score (es): 0. Result ID: 69d4b9af47f34ce924bc5707.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 1B"}. This record does not assert independent reproduction.

### LDA

Score (es): 0. Result ID: 69d4b9af47f34ce924bc5708.

[Reporting paper](<https://www.sota2.com/research/paper/69d48fa67d0172140ec74423.md>)

Recorded evaluation configuration: {"backbone":"Gemma 3 1B"}. This record does not assert independent reproduction.

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