List your company

Long-Context Multi-Hop Evaluation on HotpotQA Long (2.1k tok)

43.3Exact Match (EM)TUR-DPO
34.35636.6783941.322Apr 30, 2026

Evaluation results

MethodLinks
TUR-DPO2026.0443.366.18.7
DPO2026.0439.558.612.4
SFT2026.0434.752.316.2