Negative reactivity caused by radial expansion of core is known as one of the inherent safety features in a sodium-cooled fast reactor (SFR). In order to confirm the applicability of the reactivity model, the authors have been conducting the benchmark exercises of the unprotected loss of heat sink event tests in a pool-type experimental SFR EBR-II. In the blind phase in the benchmark analyses using the in-house plant dynamics analysis (1D) code, it was found the increase of the core inlet temperature was evaluated lower than that of the measured data and the feedback reactivity was underestimated. Since the discrepancies might be brought by the ignorance of thermal stratification in the cold pool by means of the perfect mixing model in the 1D code, the detailed model of the cold pool for the computational fluid dynamics (CFD) code was introduced and the 1D-CFD codes coupling method was applied to the benchmark analyses. Then, it was confirmed that both the thermal stratification in the cold pool and the core inlet temperature were successfully reproduced. It could insist that modeling the cold pool including the IHX outlet window in detail was important to simulate the thermal hydraulics behaviors including thermal stratification there.