ResearchBenchmarksMulti-hop Reasoning on LongBench MuSiQue and WikiMultiHopQAFollow69.9F1 ScoreMGRS57.10860.42963.7567.071Nov 28, 2025Evaluation ResultsMethodMethodLinksF1 ScoreMGRSBackbone=GPT-4o-miniBackbone=GPT-4o-mini2025.1169.9AoTBackbone=GPT-4o-miniBackbone=GPT-4o-mini2025.1168.5AFlowBackbone=GPT-4o-miniBackbone=GPT-4o-mini2025.1161FoTBackbone=GPT-4o-mini, n=8Backbone=GPT-4o-mini, n=82025.1159.1CoT-SCBackbone=GPT-4o-mini, n=5Backbone=GPT-4o-mini, n=52025.1158.6Self-refineBackbone=GPT-4o-miniBackbone=GPT-4o-mini2025.1158.2CoTBackbone=GPT-4o-miniBackbone=GPT-4o-mini2025.1157.6