List your company

Multi-hop QA on MuSiQue (Violation Metrics)

0.83Avg ViolationSAVER
0.73321.38662.042.6934Apr 9, 2026

Evaluation results

MethodLinks
SAVERModel Scale=LLaMA-3.1-8B2026.040.8369.380.1119.73
MADModel Scale=LLaMA-3.1-8B2026.042.1626.17—36.51
CoTModel Scale=LLaMA-3.1-8B2026.042.9113.26—37.58
VANILLA LMModel Scale=LLaMA-3.1-8B2026.043.255.34—62.63