# PRP Faithfulness Evaluation on Eve 30 persona statements (original characters)

Source: [PRP Faithfulness Evaluation on Eve 30 persona statements (original characters)](<https://www.sota2.com/research/sota/prp-faithfulness-evaluation-on-eve-30-persona-statements-original-characters>)

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

Data updated: 2026-02-26T06:11:24.508Z

Evaluation of model faithfulness concerning PRP (Persona Reference Preservation) using the Eve 30 persona statements dataset, measured by ΔAPC (DeB), ΔAPC (GPT-4), and Human evaluation scores.

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 | PRP Faithfulness Evaluation |
| Dataset (including any stated split) | Eve 30 persona statements (original characters) |
| Metric | ΔAPC (DeB) |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 7 of 7 |
| 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

- [ΔAPC (DeB)](<https://www.sota2.com/research/sota/prp-faithfulness-evaluation-on-eve-30-persona-statements-original-characters.md?metric=%CE%94APC%20(DeB)>): Lower is better; unit: SCALAR.

- [ΔAPC (GPT-4)](<https://www.sota2.com/research/sota/prp-faithfulness-evaluation-on-eve-30-persona-statements-original-characters.md?metric=%CE%94APC%20(GPT-4)>): Lower is better; unit: SCALAR.

- [Human Score](<https://www.sota2.com/research/sota/prp-faithfulness-evaluation-on-eve-30-persona-statements-original-characters.md?metric=Human%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | ΔAPC (DeB) | Result date | Report paper |
| --- | --- | --- | --- |
| Gemma-7B | 0.7 | Not recorded | 68fa753e1e209131cea66b68 |
| Experience Upload (EU) | 3.6 | Not recorded | 68fa753e1e209131cea66b68 |
| Long-context Memory (LCM) | 3.9 | Not recorded | 68fa753e1e209131cea66b68 |
| Experience Upload (EU) | 3.9 | Not recorded | 68fa753e1e209131cea66b68 |
| Retrieval-augmented Generation (RAG) | 4.8 | Not recorded | 68fa753e1e209131cea66b68 |
| Long-context Memory (LCM) | 5.1 | Not recorded | 68fa753e1e209131cea66b68 |
| Retrieval-augmented Generation (RAG) | 5.8 | Not recorded | 68fa753e1e209131cea66b68 |

### Gemma-7B

ΔAPC (DeB): 0.7. Result ID: 699397123a6c314849171664.

[Reporting paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Implementation](<https://github.com/KomeijiForce/Active_Passive_Constraint_Koishiday_2024>)

Recorded evaluation configuration: {"CPO (APC-based DPO)":"False"}. This record does not assert independent reproduction.

### Experience Upload (EU)

ΔAPC (DeB): 3.6. Result ID: 699397123a6c314849171665.

[Reporting paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Implementation](<https://github.com/KomeijiForce/Active_Passive_Constraint_Koishiday_2024>)

Recorded evaluation configuration: {"CPO (APC-based DPO)":"False"}. This record does not assert independent reproduction.

### Long-context Memory (LCM)

ΔAPC (DeB): 3.9. Result ID: 699397123a6c314849171666.

[Reporting paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Implementation](<https://github.com/KomeijiForce/Active_Passive_Constraint_Koishiday_2024>)

Recorded evaluation configuration: {"CPO (APC-based DPO)":"False"}. This record does not assert independent reproduction.

### Experience Upload (EU)

ΔAPC (DeB): 3.9. Result ID: 699397123a6c314849171668.

[Reporting paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Implementation](<https://github.com/KomeijiForce/Active_Passive_Constraint_Koishiday_2024>)

Recorded evaluation configuration: {"CPO (APC-based DPO)":"True"}. This record does not assert independent reproduction.

### Retrieval-augmented Generation (RAG)

ΔAPC (DeB): 4.8. Result ID: 699397123a6c314849171667.

[Reporting paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Implementation](<https://github.com/KomeijiForce/Active_Passive_Constraint_Koishiday_2024>)

Recorded evaluation configuration: {"CPO (APC-based DPO)":"False"}. This record does not assert independent reproduction.

### Long-context Memory (LCM)

ΔAPC (DeB): 5.1. Result ID: 699397123a6c314849171669.

[Reporting paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Implementation](<https://github.com/KomeijiForce/Active_Passive_Constraint_Koishiday_2024>)

Recorded evaluation configuration: {"CPO (APC-based DPO)":"True"}. This record does not assert independent reproduction.

### Retrieval-augmented Generation (RAG)

ΔAPC (DeB): 5.8. Result ID: 699397123a6c31484917166a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753e1e209131cea66b68.md>)

[Implementation](<https://github.com/KomeijiForce/Active_Passive_Constraint_Koishiday_2024>)

Recorded evaluation configuration: {"CPO (APC-based DPO)":"True"}. This record does not assert independent reproduction.

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