The WiFi-based indoor localization problem aimsto identify the location of a user using the signals receivedfrom surrounding wireless access points. A major approach toaddress this problem is through machine learning algorithmstrained on precollected radio maps. However, these approacheseither completely ignore the temporal aspects of the problem orthe interval between consecutive reference points is too large.Therefore, in this work, we study the application of end-to-endsequence models for ﬁner-level WiFi-based indoor localization.We show that localization task can be formulated as a sequencelearning problem by using recurrent neural networks withregression output. The regression output is used to estimatethree-dimensional positions and allows the network to easilyscale to larger areas. In addition, we present our WiFine dataset containing 290 trajectories sequentially collected at ﬁner-levelreference points. The dataset is made publicly available foradvancing sequential indoor localization research. The exper-iments performed on WiFine dataset show that on ﬁner-levellocalization task the recurrent neural networks are superiorto non-sequential models such as k-nearest neighbors andfeedforward neural network.