List your company

Multi-hop Question Answering on MuSiQue (Hop-Specific F1 Breakdown)

40.62-hop F1SAKE
24.68828.81932.9537.081Feb 10, 2026

Evaluation results

MethodLinks
SAKEBackbone=Qwen3-4B-Thinking2026.0240.624.112
SAKEBackbone=QwQ-32B2026.0237.62320.4
Search-o1Backbone=QwQ-32B2026.0235.81512.4
SAKEBackbone=Qwen3-30B-A3B-Thinking2026.0235.21814.6
Search-o1Backbone=Qwen3-4B-Thinking2026.0234.918.89.58
Standard RAGBackbone=Qwen3-30B-A3B-Thinking2026.02322215.8
Standard RAGBackbone=Qwen3-4B-Thinking2026.0231.521.610.9
Standard RAGBackbone=QwQ-32B2026.0229.618.415.3
Search-o1Backbone=Qwen3-30B-A3B-Thinking2026.0225.314.85.94