List your company

Fact Consolidation (Single-Hop) on MemoryAgentBench (MAB) FC-SH 262K

93AccuracyAblation C (gpt-4o backbone)
3.5626.785073.22May 31, 2026

Evaluation results

MethodLinks
Ablation C (gpt-4o backbone)Pipeline=Ablation C (gpt-4o backbone), Backbone=gpt-4o2026.0593
HeadlinePipeline=Headline, Backbone=gpt-4o-mini2026.0582
Ablation A (chunk-4096)Pipeline=Ablation A (chunk-4096), Backbone=gpt-4o-mini2026.0573
LLM-judgment baselinePipeline=LLM-judgment baseline, Backbone=gpt-4o-mini2026.0561
GPT-4o (long-context)system=long-context2026.0560
HippoRAG-v2 (best published)2026.0554
BM25 (MAB chunk-512 default)2026.0548
Zep / Graphiti (temporal KG)2026.057