List your company

Multihop Question Answering on HotpotQA In-Distribution (Context Window Scaling)

59.4Accuracy (8K Context)MemReread
26.01634.68343.3552.017May 11, 2026

Evaluation results

MethodLinks
MemRereadScale=4B2026.0559.457.851.646.949.253.151.654.753
ReMemR1Scale=4B2026.0558.654.7575051.645.35052.352.4
MemAgentScale=4B2026.0553.955.546.950.843.84343.842.247.5
MemAgentScale=1.7B2026.0544.539.129.723.42520.321.119.527.8
MemRereadScale=1.7B2026.0543433223.422.722.723.417.228.4
ReMemR1Scale=1.7B2026.0527.328.131.219.525.825.821.917.224.6