List your company

Single-Choice Question Answering on RefMem-Bench

66.2AccuracyREMIND
44.15249.87655.661.324May 31, 2026

Evaluation results

MethodLinks
REMINDBackbone=Qwen3-VL-8B2026.0566.252.3
Qwen3-VL-30BBackbone=30B2026.0558.340.5
LangMemBackbone=GPT-4o-mini2026.0557.8—
A-MemBackbone=GPT-4o-mini2026.0557.444.2
LightMemBackbone=GPT-4o-mini2026.0556.944.1
GAMBackbone=GPT-4o-mini2026.0556.441.4
DS-Distill-Qwen-14BBackbone=14B2026.0554.945.7
MemAgent-7BBackbone=7B2026.0553.543.9
MemoryOSBackbone=GPT-4o-mini2026.0553.140.9
MEM1-7BBackbone=7B2026.0549.236.9
Mem0Backbone=GPT-4o-mini2026.0548.5—
MemGPTBackbone=GPT-4o-mini2026.0547.739.3
- w/ NaiveRAGBackbone=Qwen3-VL-8B2026.0547.638.6
Base ModelBackbone=Qwen3-VL-8B2026.054537.8