ResearchBenchmarksFidelity Evaluation on BDI/E Dialogue DatasetFollow93B-MQwen3-14B36.8451.426680.58Jun 28, 2026Evaluation ResultsMethodMethodLinksB-MD-MI-MB-1Dist-1Qwen3-14Bnumber of parameters=14Bnumber of parameters=14B2026.0693805111.295.3GPT-5.52026.0681792713.396.2Dual-LLM2026.0681592815.896DSv4-Pro2026.0679663217.296.5CogWM-14Bnumber of parameters=14Bnumber of parameters=14B2026.0639261432.294.8