A dataset of simultaneous two-photon calcium imaging and auditory discrimination behavior

The emerging field of artificial intelligence for neuroscience relies heavily on high-quality, standardized datasets. Data linking high-dimensional neural population dynamics with complex behavior is crucial for advancing the field of brain-computer interfaces. However, few existing datasets provide these paired modalities in a standardized, AI-ready format, limiting their immediate utility for computational modeling. Here, we present a large-scale dataset comprising simultaneous two-photon calcium imaging of the primary auditory cortex (A1) and detailed behavioral records in mice. The animals performed an auditory two-alternative forced-choice (2AFC) categorization task, classifying tones as low or high frequency by licking left or right water ports. The raw neural and behavioral data were preprocessed, synchronized, and formatted into trial-aligned multi-dimensional tensors. Comprehensive validation confirms that the recorded populations exhibit consistent, high-fidelity temporal dynamics and distinct frequency tuning across all experimental conditions. To evaluate the dataset’s capacity for neural decoding, we benchmarked it across diverse machine learning architectures and deep learning networks. By bridging biological recordings and behavior, this open-access dataset serves as a valuable benchmark for developing novel decoding algorithms and testing brain-inspired computational models.