List your company

Multi-hop QA on 2WikiMHQA

0.56Average ViolationSAVER
0.46921.08211.6952.3079Apr 9, 2026

Evaluation results

MethodLinks
SAVERModel Scale=LLaMA-3.1-8B2026.040.5672.34813.84
MADModel Scale=LLaMA-3.1-8B2026.041.8132.78—28.82
CoTModel Scale=LLaMA-3.1-8B2026.042.2117.41—32.11
VANILLA LMModel Scale=LLaMA-3.1-8B2026.042.836.58—53.19