Long-context Question Answering on LoCoMo Single-Hop 2024
48.66F1 ScoreLyzr Cognis
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Lyzr CognisBackbone=GPT-4.12026.03 | 48.66 | 35.76 | 71.99 | |
| Mem0Backbone=GPT-4.12026.03 | 38.72 | 27.13 | 67.13 | |
| Mem0gBackbone=GPT-4.1, Variant=graph-based knowledge representation2026.03 | 38.09 | 26.03 | 65.71 | |
| ZepBackbone=GPT-4.12026.03 | 35.74 | 23.3 | 61.7 | |
| LangMemBackbone=GPT-4.1, Framework=LangChain-based2026.03 | 35.51 | 26.86 | 62.23 | |
| OpenAIBackbone=GPT-4.12026.03 | 34.3 | 23.72 | 63.79 | |
| A-MemBackbone=GPT-4.1, Management=structured2026.03 | 27.02 | 20.09 | — | |
| MemGPTBackbone=GPT-4.1, Paging=OS-inspired2026.03 | 26.65 | 17.72 | — | |
| LoCoMoBackbone=GPT-4.12026.03 | 25.02 | 19.75 | — | |
| A-Mem*Backbone=GPT-4.1, Evaluation=LLM-as-a-Judge2026.03 | 20.76 | 14.9 | 39.79 | |
| ReadAgentBackbone=GPT-4.1, Memory=gist-based2026.03 | 9.15 | 6.48 | — | |
| MemoryBankBackbone=GPT-4.1, Memory=consolidation2026.03 | 5 | 4.77 | — |