AbstractThe lack of suitable benchmark problems makes it difficult to test sensitivity codes with a covariance library. A benchmark problem has therefore been defined for one- and two-dimensional sensitivity and uncertainty analysis codes and code systems. The problem, representative of a fusion reactor blanket, has a simple, three-zone, r — z geometry containing a D—T fusion neutron source distributed in a central void region surrounded by a thick 6LiH annulus. The response of interest is the 6Li tritium production per source neutron, T6. The calculation has been performed with SENSIBL using other codes from the AARE code system as a test of both SENSIBL and the linked, modular system. The calculation was performed using the code system in the standard manner with a covariance data library in the COVFILS-2 format but modified to contain specifically tailored covariance data for H and 6Li (Path A). The calculation was also performed by a second method which uses specially perturbed H and Li cross sections (Path B). This method bypasses SENSIBL and allows a hand calculation of the benchmark T6 uncertainties. The results of Path A and Path B were total uncertainties in T6 of 0.21% and 0.19%, respectively. The closeness of the results for this challenging test gives confidence that SENSIBL and the AARE system will perform well for realistic sensitivity and uncertainty analyses.