A new method for stochastic unraveling of general time-local quantum master equations (QME) which involve the reduced density operator at time t only is proposed. The present kind of jump algorithm enables a numerically efficient treatment of QMEs which are not of Lindblad form. So it opens new large fields of application for stochastic methods. The unraveling can be achieved by allowing for trajectories with negative weight. We present results for the quantum Brownian motion and the Redfield QMEs as test examples. The algorithm can also unravel non-Markovian QMEs when they are in a time-local form like in the time-convolutionless formalism. 1 for a number of typical examples). These QMEs describe the time evolution of density matrices which are used in order to represent the mixed nature of the states. Stochastic unraveling is an efficient numerical tool for solving such equations. This method allows one to simulate much larger and more complex systems with many degrees of freedom. It can, for example, be used to accurately describe femtochemical experiments in the liquid phase whose description has been limited until now to models with one or two effective interaction coordinates. In the unraveling scheme one considers an ensemble of stochastic Schrödinger equations (SSEs) which in the limit of a large ensemble resembles the respective QME. The numerical effort scales much more favorably with the size of the basis since one is now dealing with wave functions and not density matrices anymore (for a comparison of direct integrators, see Ref. 2). Another aspect of the stochastic methods is the possible physical interpretation of experiments detecting macroscopic fluctuations (e.g. photon counting) in various quantum systems [3]. Most of the unraveling schemes [3,4,5,6,7,8] have been restricted to QMEs of Lindblad form [9] which ensures that the reduced density matrix (RDM) stays positive semi-definite for all times and all parameters. Nevertheless there are many physical meaningful QMEs which result in positive-or almost positive-definite RDMs although they are not of Lindblad form. The increasing interest in descriptions beyond the Lindblad class such as the quantum Brownian motion [10,11] developed a method how to exactly represent the RDM of a system coupled to a linear heat bath in terms of SSEs. The numerical properties of this approach need to be explored. Breuer et al.[15] extended a scheme which they had used to calculate multi-time correlation functions [19] to the unraveling of QMEs. Their technique is based on doubling the Hilbert space. Instead of a single stochastic wave function one has a pair of them [15]. This scheme conserves Hermiticity of the RDM only on average and not for every single realization. Thus, the deviation from Hermiticity is a quantity with statistical error and one has to perform a huge number of realizations in order to achieve good convergence. Since stability and efficiency are crucial issues for unraveling algorithms we propose in this article an alternative approach which fu...