List your company

Question Answering on LoCoMo excluding adversarial (test)

55.5Overall F1MemRouter
7.1419.69532.2544.805May 1, 2026

Evaluation results

MethodLinks
MemRouterAnswer Agent / Backbone=Qwen3.5-35B-A3B, Backbone Transfer Setting=same router, no router retraining, Retrieval and Prompting Pipeline=Own system-level2026.0555.560.956.149.723.8
MemRouterAnswer Agent / Backbone=Qwen2.5-7B-Instruct, Retrieval and Prompting Pipeline=Own system-level2026.055257.552.44427.1
Memory-R1-GRPOAnswer Agent / Backbone=Qwen2.5-7B-Instruct, Retrieval and Prompting Pipeline=Own system-level2026.0543.133.623.647.846.9
MemoryOSAnswer Agent / Backbone=Qwen2.5-7B-Instruct, Retrieval and Prompting Pipeline=Own system-level2026.0534.629.62126.340.9
Mem0Answer Agent / Backbone=Qwen2.5-7B-Instruct, Retrieval and Prompting Pipeline=Own system-level2026.0530.62520.333.232.7
A-MemAnswer Agent / Backbone=Qwen2.5-7B-Instruct, Retrieval and Prompting Pipeline=Own system-level2026.0526.11914.723.730.6
LoCoMo (RAG)Answer Agent / Backbone=Qwen2.5-7B-Instruct, Retrieval and Prompting Pipeline=Own system-level2026.0599.611.88.48.7