List your company

Multi-Hop Fact-based Reasoning on MAB FC-MH, 262K v3 (test)

27AccuracyChain-Aware Resolution (CAR)
2.048.521521.48May 31, 2026

Evaluation results

MethodLinks
Chain-Aware Resolution (CAR)Backbone=gpt-4o-mini, Pipeline architecture=multi-hop2026.0527
GPT-4o (long-context)Backbone=gpt-4o, Pipeline architecture=long-context2026.055
HippoRAG-v2 (best published)Backbone=gpt-4o-mini2026.055
BM25Backbone=gpt-4o-mini, Pipeline architecture=MAB chunk-512 default2026.053
Zep / GraphitiBackbone=gpt-4o-mini, Pipeline architecture=temporal KG2026.053