List your company

Question Answering on 2-Hop QA Parametric Knowledge

43.07AccuracyCoT
31.47434.484537.49540.5055Feb 9, 2026

Evaluation results

MethodLinks
CoTBase Model=Qwen2.5-Base-1.5B2026.0243.071
Thinking StatesBase Model=Qwen2.5-Base-1.5B2026.0243.051.23
iCoTBase Model=Qwen2.5-Base-1.5B2026.0236.311.8
CoconutBase Model=Qwen2.5-Base-1.5B2026.0232.61.21
No CoTBase Model=Qwen2.5-Base-1.5B2026.0231.922.03