In order to check the real instrumental performance and assess the feasibility of the science goals, the science working groups of LAMOST came up with the pilot survey of LAMOST, which was developed before the regular spectral survey. The pilot survey was launched on Oct 24 2011, and ended on June 2012, which includes nine full moon cycles and basically covers all good observing seasons in a year at the site. Pilot data release——To provide useful data to public, low quality data is cut out in this early data release. To publish relatively high quality spectra from all the data, the first selection criteria is SNR. After cutting the SNR≤10 (smaller than 10 both in g and r band), 493,000 spectra are kept. As we know, SNR of a spectra is correlated with magnitude if exposure time and other observation condition (for example, seeing) is given. We can get the theoretical relation between SNR and magnitude, comparing it with a fitting relationship from each 4000 spectra taken at the mean time. Then, the relationship is fixed, and those spectra with SNR deviate far from the relationship should be cut. After this cutting, 476,000 spectra kept. Another point we should consider is about the pointing of the fibers, if a given fiber pointed to an exact object that we want to. We calculated repeat precision of each fiber based on the SNR magnitude, and find out ‘good‘fibers with repeat precision higher than 95%. We cut those spectra observed through ‘bad‘fibers, and 319,000 spectra are left as the released data.