There is no such axiom of science. Actually, I think you can't claim that science "describes" the world we perceive, nor that science has a "goal".
A goal is a social construct, generally adopted with reference to personal objectives, which are inevitably social. Scientists work hard to make scientific progress in order to reach personal goals, not scientific goals.
Scientific knowledge is the ability to predict significant outcomes of events occurring in time. The ability to predict the outcome of events has economic and/or social value, and so confers power on those who have it.
The ability to predict outcomes is achieved by producing a model (or "algorithm", or "theory", or "law", or "principle" or whatever you want to call it) which can be used to predict, given certain preconditions, the occurrence of subsequent likely events and outcomes over time. You might think a model is "descriptive" of reality but it is not - it is only predictive of outcomes. The descriptive element is merely a by-product of the ability to predict, which is the important part. For example, Newton's law of gravity is a model that predicts the behaviour of physical bodies by mathematically modelling the influence of an attractive force between them (which depends on their mass and distance apart etc.). Newton's "law" is really very good at predicting the motions of planets around the sun. However it does not describe the planets themselves or say anything about how this gravitational force is created or sustained. The model does not say "gravity exists". It just says "by using a model of an attractive force between bodies, let's call that 'gravity', we get good predictions of how those bodies will move".
A model is only "useful" if it sufficiently reliably predicts outcomes that have some social significance, desirable or undesirable, and hence have value. This gives the scientist the means to achieve their objectives. The scientist (or others paying the scientist), being empowered by using the scientific model, can then take action that makes (economic or social) use of the predictions in some way. The degree of reliability or accuracy of the model's predictions determines how useful the model is.
Scientific progress by humanity is measured only in terms of the ability of scientific models to predict outcomes previously unpredictable, or more accurately to predict outcomes - that is, if you are able to make predictions of outcomes more accurately than previously possible and this increased accuracy makes the model more useful, then you have made some progress.
The predictive power of a scientific model may also be useful in reverse - that is, starting from observed outcomes, to deduce what must have been the preconditions. For example the science of kinetics can be used to predict the outcomes of interactions between moving bodies, but also to deduce the precursors, and possibly causes, of collisions between bodies by observing the outcomes. Investigators do this when trying to establish the sequence of events preceding accidents.
The "brain in a vat" possibility is not really relevant to the issue. It does not matter whether or not our senses perceive "reality", just like it does not matter if scientific models do or do not describe "reality". Philosophers might be interested in musing on whether, if we knew that we are all brains in vats, our perceptions of ourselves and our moral behaviour would be different in some way, but that's not really a scientific question. Science is interested only if a "brain in a vat" model is better at predicting outcomes than other theories of the universe.
As a philosopher you might wonder whether there will be a future in which humans are able perfectly to predict every outcome of every event. But currently, as far as I am aware, we have no scientific model that could predict such a future. It certainly is not an axiom of science.