Zoom: https://stanford.zoom.us/my/sihanyuan?pwd=QnpsUHZWWGJ2ekVYWmZVL3BmM0gzZ…

Galaxy surveys have advanced from photometric plates operated by human to spectroscopic fibers positioned by robots. Meanwhile, analysis methods of galaxy clustering still largely rely on the two- and three-point correlation functions, or the power spectrum and bispectrum. This leaves open the question of whether and how much information can be (robustly) extracted from ongoing and upcoming galaxy clustering data. In this talk, I will first introduce a Lagrangian, EFT-based, forward-modeling framework (LEFTfield) to perform field-level Bayesian inference (FBI) of galaxy clustering. FBI aims to extract constraints on cosmological parameter(s) from the data directly at the field level, i.e. at the level of the three-dimensional map of galaxies. Using constraints on the amplitude of matter fluctuations (sigma8) as the metric, I will further present the first direct, apple-to-apple comparison between FBI and an inference using the combination of power spectrum plus bispectrum (P+B)—both with exactly the same forward model and analysis scale cut—on dark matter halos in N-body simulations. The FBI constraints show an improvement of 3.5 (5.2) times over the P+B constraints, even at the modest analysis cutoff scale of kmax=0.1h/Mpc (0.12 h/Mpc). These results underline the wealth of cosmological information in nonlinear galaxy clustering, currently beyond the reach of low-order n-point functions. Finally, I will conclude with some results from a public data challenge, the ``Beyond-2pt parameter-masked data challenge’’. This challenge sets in motion a community effort to validate and showcase novel statistics and methods to analyze data from upcoming galaxy surveys, extending beyond the canonical two-point statistics.