Dementia Care Dialogue Evaluation

0.86GMCPQT2-GRPO

Evaluation results

Method
T2-GRPO2026.06	0.86	0.83	0.78	0.82	0.81	0.84	0.83	2.1
Full T2-GRPO2026.06	0.86	0.83	0.78	0.82	0.81	0.84	0.83	2.1
GDPO + Rank norm. + Dual horizon2026.06	0.85	0.84	0.77	0.82	0.83	0.82	0.83	6.8
GDPO + Rank norm.2026.06	0.83	0.73	0.75	0.77	0.78	0.75	0.77	7.4
GDPO + Rank norm. + Safety veto2026.06	0.82	0.72	0.74	0.76	0.77	0.76	0.77	2.4
GDPO + Dual horizon2026.06	0.8	0.81	0.74	0.78	0.76	0.74	0.75	7.1
GDPO + Dual horizon + Safety veto2026.06	0.8	0.82	0.73	0.78	0.76	0.75	0.76	2.2
GDPO2026.06	0.79	0.71	0.73	0.74	0.75	0.72	0.74	7.6
GDPO2026.06	0.79	0.71	0.73	0.74	0.75	0.72	0.74	7.6
GDPO + Safety veto2026.06	0.77	0.69	0.7	0.72	0.74	0.7	0.72	2.8
GPT-5.42026.06	0.75	0.62	0.6	0.66	0.82	0.85	0.84	8.9
Gemini 3.12026.06	0.73	0.58	0.64	0.65	0.84	0.77	0.81	9.4
GRPO2026.06	0.7	0.63	0.68	0.67	0.71	0.68	0.7	11.5
Qwen3.5-122B-A10B2026.06	0.68	0.51	0.56	0.58	0.78	0.81	0.8	13.8
PPO2026.06	0.63	0.56	0.65	0.61	0.65	0.62	0.64	12.3
SFT only2026.06	0.6	0.53	0.58	0.57	0.68	0.65	0.67	14.1
Qwen3.5-9B2026.06	0.52	0.41	0.48	0.47	0.62	0.58	0.6	22.4