A series of benchmark calculations of critical experiments has been performed to assess the effects that recent changes in the ENDF/B data files have had on calculated liquid-metal fast breeder reactor parameters. Three well-documented critical assemblies were studied using standard methods of fast reactor analysis (two-dimensional multigroup diffusion theory) with both ENDF/B-III and -IV. A review of the changes in the principal cross sections incorporated in the latest evaluation has been made and was used to interpret the changes in calculated integral parameters. Some of the principal cross-section modifications included in ENDF/B-IV were: decreasing (≈2%),  src ="fig002.gif" alt=""/> (≈1 to 1.5%), and  src ="fig003.gif" alt=""/>  (≈3 to 4%) and increasing  src ="fig004.gif" alt=""/> (≈1 to 1.5%). In general, improved agreement between measured and calculated integral parameters has been obtained with the new ENDF/B evaluation. The overprediction of 28c/49 ƒ is reduced slightly (≈1.5%); the central worth calculation-to-experiment discrepancy is generally reduced (≈5 to 15%); and little change occurs in the calculated eigenvalues that remain ≈1.0 to 1.5% δk low.