Persona Memory Management for Language Agents on PersonaMem 128k
63.87AccuracyRGMem
Evaluation Results
| Method | Links | |
|---|---|---|
| RGMemInput Tok. (k)=1,421.79, Output Tok. (k)=240.33, Runtime (s)=5,441.02, API Calls=978.54, Underlying LLM=gpt-4, Fine-tuning Protocol=None2025.10 | 63.87 | |
| Mem0Input Tok. (k)=1,336.24, Output Tok. (k)=215.12, Runtime (s)=5,507.65, API Calls=1,029.22, Underlying LLM=gpt-4, Fine-tuning Protocol=None2025.10 | 56.79 | |
| MemoryOSInput Tok. (k)=3,489.71, Output Tok. (k)=480.61, Runtime (s)=9,125.66, API Calls=3,278.42, Underlying LLM=gpt-4, Fine-tuning Protocol=None2025.10 | 54.23 | |
| LangMemInput Tok. (k)=1,128.17, Output Tok. (k)=134.51, Runtime (s)=2,719.02, API Calls=632.51, Underlying LLM=gpt-4, Fine-tuning Protocol=None2025.10 | 52.36 | |
| A-MemInput Tok. (k)=1,712.45, Output Tok. (k)=357.31, Runtime (s)=6,312.21, API Calls=1,654.32, Underlying LLM=gpt-4, Fine-tuning Protocol=None2025.10 | 49.17 |