# Dementia Care Dialogue Evaluation

Source: [Dementia Care Dialogue Evaluation](<https://www.sota2.com/research/sota/dementia-care-dialogue-evaluation>)

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

Data updated: 2026-06-09T09:54:04.828Z

Evaluation of dialogue systems specifically designed for dementia care scenarios, utilizing metrics such as GMCPQ, PACES, and PCCBP to assess quality and safety.

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 | Dementia care dialogue |
| Dataset (including any stated split) | Dementia care dialogue |
| Metric | GMCPQ |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 17 of 17 |
| 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

- [GMCPQ](<https://www.sota2.com/research/sota/dementia-care-dialogue-evaluation.md?metric=GMCPQ>): Higher is better; unit: SCALAR.

- [PACES](<https://www.sota2.com/research/sota/dementia-care-dialogue-evaluation.md?metric=PACES>): Higher is better; unit: SCALAR.

- [PCCBP](<https://www.sota2.com/research/sota/dementia-care-dialogue-evaluation.md?metric=PCCBP>): Higher is better; unit: SCALAR.

- [CQ Avg.](<https://www.sota2.com/research/sota/dementia-care-dialogue-evaluation.md?metric=CQ%20Avg.>): Higher is better; unit: SCALAR.

- [Naturalness](<https://www.sota2.com/research/sota/dementia-care-dialogue-evaluation.md?metric=Naturalness>): Higher is better; unit: SCALAR.

- [Authenticity](<https://www.sota2.com/research/sota/dementia-care-dialogue-evaluation.md?metric=Authenticity>): Higher is better; unit: SCALAR.

- [DQ Avg.](<https://www.sota2.com/research/sota/dementia-care-dialogue-evaluation.md?metric=DQ%20Avg.>): Higher is better; unit: SCALAR.

- [Violations Count](<https://www.sota2.com/research/sota/dementia-care-dialogue-evaluation.md?metric=Violations%20Count>): Lower is better; unit: SCALAR.

## Recorded results

| Method | GMCPQ | Result date | Report paper |
| --- | --- | --- | --- |
| T2-GRPO | 0.86 | Not recorded | 6a279e7690966ade919bfb6d |
| Full T2-GRPO | 0.86 | Not recorded | 6a279e7690966ade919bfb6d |
| GDPO + Rank norm. + Dual horizon | 0.85 | Not recorded | 6a279e7690966ade919bfb6d |
| GDPO + Rank norm. | 0.83 | Not recorded | 6a279e7690966ade919bfb6d |
| GDPO + Rank norm. + Safety veto | 0.82 | Not recorded | 6a279e7690966ade919bfb6d |
| GDPO + Dual horizon | 0.8 | Not recorded | 6a279e7690966ade919bfb6d |
| GDPO + Dual horizon + Safety veto | 0.8 | Not recorded | 6a279e7690966ade919bfb6d |
| GDPO | 0.79 | Not recorded | 6a279e7690966ade919bfb6d |
| GDPO | 0.79 | Not recorded | 6a279e7690966ade919bfb6d |
| GDPO + Safety veto | 0.77 | Not recorded | 6a279e7690966ade919bfb6d |
| GPT-5.4 | 0.75 | Not recorded | 6a279e7690966ade919bfb6d |
| Gemini 3.1 | 0.73 | Not recorded | 6a279e7690966ade919bfb6d |
| GRPO | 0.7 | Not recorded | 6a279e7690966ade919bfb6d |
| Qwen3.5-122B-A10B | 0.68 | Not recorded | 6a279e7690966ade919bfb6d |
| PPO | 0.63 | Not recorded | 6a279e7690966ade919bfb6d |
| SFT only | 0.6 | Not recorded | 6a279e7690966ade919bfb6d |
| Qwen3.5-9B | 0.52 | Not recorded | 6a279e7690966ade919bfb6d |

### T2-GRPO

GMCPQ: 0.86. Result ID: 6a27e296fbc373592c25212e.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Main results"}. This record does not assert independent reproduction.

### Full T2-GRPO

GMCPQ: 0.86. Result ID: 6a27e296fbc373592c252136.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Ablation study final"}. This record does not assert independent reproduction.

### GDPO + Rank norm. + Dual horizon

GMCPQ: 0.85. Result ID: 6a27e296fbc373592c252133.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Pairwise combination ablation"}. This record does not assert independent reproduction.

### GDPO + Rank norm.

GMCPQ: 0.83. Result ID: 6a27e296fbc373592c252130.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Ablation study component"}. This record does not assert independent reproduction.

### GDPO + Rank norm. + Safety veto

GMCPQ: 0.82. Result ID: 6a27e296fbc373592c252134.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Pairwise combination ablation"}. This record does not assert independent reproduction.

### GDPO + Dual horizon

GMCPQ: 0.8. Result ID: 6a27e296fbc373592c252131.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Ablation study component"}. This record does not assert independent reproduction.

### GDPO + Dual horizon + Safety veto

GMCPQ: 0.8. Result ID: 6a27e296fbc373592c252135.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Pairwise combination ablation"}. This record does not assert independent reproduction.

### GDPO

GMCPQ: 0.79. Result ID: 6a27e296fbc373592c25212d.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69608b7953fdd5c361744490.md>)

[Implementation](<https://github.com/NVlabs/GDPO>)

Recorded evaluation configuration: {"Setup":"Main results"}. This record does not assert independent reproduction.

### GDPO

GMCPQ: 0.79. Result ID: 6a27e296fbc373592c25212f.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69608b7953fdd5c361744490.md>)

[Implementation](<https://github.com/NVlabs/GDPO>)

Recorded evaluation configuration: {"Setup":"Ablation study baseline"}. This record does not assert independent reproduction.

### GDPO + Safety veto

GMCPQ: 0.77. Result ID: 6a27e296fbc373592c252132.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Ablation study component"}. This record does not assert independent reproduction.

### GPT-5.4

GMCPQ: 0.75. Result ID: 6a27e296fbc373592c252126.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Main results"}. This record does not assert independent reproduction.

### Gemini 3.1

GMCPQ: 0.73. Result ID: 6a27e296fbc373592c252127.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Main results"}. This record does not assert independent reproduction.

### GRPO

GMCPQ: 0.7. Result ID: 6a27e296fbc373592c25212c.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73101e209131cea5a395.md>)

[Implementation](<https://github.com/deepseek-ai/DeepSeek-Math>)

Recorded evaluation configuration: {"Setup":"Main results"}. This record does not assert independent reproduction.

### Qwen3.5-122B-A10B

GMCPQ: 0.68. Result ID: 6a27e296fbc373592c252128.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Main results"}. This record does not assert independent reproduction.

### PPO

GMCPQ: 0.63. Result ID: 6a27e296fbc373592c25212b.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa63a01e209131ce9ffdaa.md>)

Recorded evaluation configuration: {"Setup":"Main results"}. This record does not assert independent reproduction.

### SFT only

GMCPQ: 0.6. Result ID: 6a27e296fbc373592c25212a.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Main results"}. This record does not assert independent reproduction.

### Qwen3.5-9B

GMCPQ: 0.52. Result ID: 6a27e296fbc373592c252129.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e7690966ade919bfb6d.md>)

Recorded evaluation configuration: {"Setup":"Main results"}. This record does not assert independent reproduction.

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