What shapes the bright end of the UV luminosity function? Stellar mass, star-formation variability, and dust in JADES

JWST surveys find more UV-bright galaxies at $z>10$ than galaxy-formation models predict, but the roles of recent star formation, stellar mass and dust remain uncertain. We analyse 1365 spectroscopically confirmed JADES galaxies at $3\leq z\leq9$, fitting NIRCam photometry and NIRSpec emission lines with \ceri. Our model includes flexible star-formation histories (SFHs) under three priors, two-component dust attenuation and nebular emission. We account for spectroscopic selection through population reweighting and assess the reliability of inferred trends. At $z=3$--5, rising SFHs are more common in low-mass galaxies, whereas most galaxies above $10^{9.5}\,{\rm M}_{\odot}$ have declining SFHs. The apparent increase in rising fractions with redshift cannot be distinguished from the adopted SFH prior alone. At fixed stellar mass, recent star formation is the strongest predictor of UV brightness. UV-brighter galaxies have higher recent SFRs, younger stellar populations and less attenuation at fixed SFR. Below $10^9\,{\rm M}_{\odot}$, they are also more likely to have rising SFHs, with odds increasing by up to a factor of 2.3 per magnitude brighter, although the strongest trend requires caution. In contrast, galaxies with $M_{\mathrm{UV}}<-20$ at $z=3$--5 are predominantly massive, and approximately half have declining SFHs. Dust partly offsets burst-driven UV brightening: attenuation increases with recent SFR and declines gradually after the inferred SFR peak. In model sensitivity tests, dust has the largest effect on bright-end counts, whereas removing galaxy-to-galaxy SFH scatter changes them by at most 0.17~dex. Thus, bursts can make low-mass galaxies unusually UV-bright for their mass, while the absolute bright end of the UVLF is primarily shaped by stellar mass and dust attenuation.