Not yet reviewed by Pith; the record is open.

The cosmic evolution of gamma-ray burst (GRB) luminosity is essential for revealing the GRB physics and for using GRBs as cosmological probes. We investigate the luminosity evolution of long GRBs with a large sample of 258 {\em Swift}/BAT GRBs. Parameterized the peak luminosity of individual GRBs evolves as $L_{\rm p}\propto{\rm }(1+z)^{k}$, we get $k=1.49\pm0.19$ using the non-parametric $\tau$ statistics method without considering observational biases of GRB trigger and redshift measurement. By modeling these biases with the observed peak flux and characterizing the peak luminosity function of long GRBs as a smoothly broken power-law with a break that evolves as $L_{\rm b}\propto (1+z)^{k_{\rm b}}$, we obtain $k_{\rm b}=1.14^{+0.99}_{-0.47}$ through simulations based on assumption that the long GRB rate follows the star formation rate (SFR) incorporating with cosmic metallicity history. The derived $k$ and $k_b$ values are systematically smaller than that reported in previous papers. By removing the observational biases of the GRB trigger and redshift measurement based on our simulation analysis, we generate mock {\em complete} samples of 258 and 1000 GRBs to examine how these biases affects on the $\tau$ statistics method. We get $k=0.94\pm 0.14$ and $k=0.80\pm 0.09$ for the two samples, indicating that these observational biases may lead to overestimate the $k$ value. With the large uncertain of $k_b$ derived from our simulation analysis, one even cannot convincingly argue a robust evolution feature of the GRB luminosity.