Understanding the mechanisms of statistical learning, where organisms infer structure from sensory observations without explicit instruction, is crucial in comprehending how the brain processes and responds to complex and dynamic environments. This work explores the innate capabilities of mice to discern environmental regularities, focusing on the role of the hippocampus in this unsupervised learning process. The study leverages a novel comparative behavioural paradigm, termed X-detection, to manipulate the statistics of embedded tone sequences (patterns) in an auditory cover task: subjects should respond to a target sound, X, that appears at a random time. We show pupil responses in mice, similar to our human subjects, track statistics of presented patterns, with limited exposure. Importantly, these embedded patterns were decoupled from the X; therefore, from the reward. Importantly, performance on X-detection was not affected by the embedded patterns. This paradigm provides a flexible and robust toolkit for probing statistical learning over a range of sound statistics without reinforcing the learning via reward. Bilateral inactivation of CA1 region of dorsal hippocampus (dCA1) despite sparing the performance on X-detection, diminished the pupil response tracking the statistics. Finally, neural recordings from dCA1 show distinct representations of task-irrelevant stimuli that reflect the tracking of the event statistics and learning abstract patterns. To our knowledge, these results provide the first set of evidence, in rodents, for causal role of hippocampus in fast statistical learning of sound statistics.