List your company

Question Answering on Crafter four-actor rollout robustness

59.3Average Exact Match (EM)S3Mem
31.16838.471545.77553.0785Apr 10, 2026

Evaluation results

MethodLinks
S3Mem2026.0459.3148.6454.461.82
Graph-NoReader2026.0457.421,920.4853.5760.68
LightMem-adapted2026.0456.37111.9453.8561.98
A-MEM-inspired2026.0452.7142.5346.4358.4
MemoryOS-adapted2026.0451.23143.9248.3553.72
Vanilla RAG2026.0446.43788.4642.5849.57
No-Memory2026.0432.25462.7331.0532.78