To obtain a full survival curve (basically a full life table) you would need to make quite a number of assumptions, since the data for paleolithic societies is scattered and probably biased. This is however not a field I know well, but I can point you to one recent paper (White, 2014) that explores how the ratio of young and old individuals in a population relates to demographic rates, using agent-based models. Using such a model you can obtain smooth survival curves and also explore how different assumptions will affect the survival curves.
There are also data from recent or current hunter-gatherer societies that can be used to approximate what the survival curves looked like during the paleolithic. For instance, Hewlett (1991) (pdf here) has compiled survival rate data for individual hunter-gatherer societies (along with other demographic rates and data from other types of pre-industrial societies). The child mortality rate (deaths before age 15) often lie in the range 40-50% in active (mobile) and sedentary hunter-gatherer societies, with an infant mortality rate usually between 10-30% (death before 12 months of age). This indicates a fairly steep survival curve, which is the reason why the average age of paleolithic people is often estimated to ~25 years. This doesn't mean that most people died around the age of 25, just that the distribution of ages was very skewed and non-normal. Many died young, but people that survived to adulthood could become much older than 25.