This dataset contains error annotations used in our IEEE RA-L paper titled 'SEDMamba: Enhancing Selective State Space Modelling with Bottleneck Mechanism and Fine-to-Coarse Temporal Fusion for Efficient Error Detection in Robot-Assisted Surgery'.We consider 24 error types for 48 videos from the SAR-RARP50 dataset. Each frame is labelled as either 'normal' (0) if no error is present or 'error' (1) if any error type occurs. It contains spatial embedding sequences, binary error annotations, and corresponding frame names at the frame level, sampled at 5Hz. For detailed implementation, please refer to the SEDMamba code.If you use this error annotation dataset, please cite the SEDMamba paper.