In this paper some improvements on certainty factor model are discussed aiming at:1)including,in a rule“E→H”,not only the CF of H when E exists but also CF of(?)when E does notexist(partly or completely).For this purpose another factor(?)is added into the original model;2)improving the model so that it can tackle problems with unknown evidence.In this aspect two conceptsare introduced:(relative)maximum existence risk and(relative)maximum non-existence risk.An impor-tant result is that even if some necessary evidence is unknown one can still know the tendency whetherthe conclusion is true.Based on the improvements a conflict resolution model for problem-level conflictis also presented In this paper some improvements on certainty factor model are discussed aiming at: 1) including, in a rule “E → H”, not only the CF of H when E exists but also CF of (?) When E does notexist (partly or For this purpose another factor (?) is added into the original model; 2) improving the model so that it can tackle problems with unknown evidence. In this aspect two conceptsare introduced: (relative) maximum occurring risk and (relative) maximum non-existence risk.An impor-tant result is that even if some necessary evidence is unknown one can still know the tendency whetherthe conclusion is true.Based on the improvements a conflict resolution model for problem-level conflictis also presented