Cross-session memory recall on CPCD-Bench
4.513LLM-as-a-Judge ScoreGPT-5.4
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GPT-5.4Model Type=General2026.05 | 4.513 | 4.448 | |
| Gemini-3-flash-previewModel Type=General2026.05 | 4.445 | 4.334 | |
| DeepSeek V3.2Model Type=General2026.05 | 4.318 | 4.208 | |
| Gemini-2.5-flashModel Type=General2026.05 | 4.296 | 4.211 | |
| Qwen3-MaxModel Type=General2026.05 | 4.205 | 4.285 | |
| GLM-5.1Model Type=General2026.05 | 4.202 | 4.189 | |
| Kimi-k2.5Model Type=General2026.05 | 4.103 | 4.125 | |
| MiniMax M2.7Model Type=General2026.05 | 4.102 | 4.012 | |
| CPCD-Chat-8BModel Type=Ours2026.05 | 3.947 | 3.944 | |
| PsyLLMModel Type=Domain-specific, Backbone=Qwen3-8B2026.05 | 3.922 | 3.899 | |
| CamelModel Type=Domain-specific, Backbone=LLaMA-8B2026.05 | 3.814 | 3.789 | |
| CBT-LLMModel Type=Domain-specific, Backbone=Baichuan-7B2026.05 | 3.811 | 3.774 | |
| CPCD-Chat-4BModel Type=Ours2026.05 | 3.756 | 3.706 | |
| Qwen3-8BModel Type=General2026.05 | 3.722 | 3.677 |