Since John S. Bell demonstrated the interest of studying linear combinations of probabilities in relation with the EPR paradox in 1964, Bell inequalities have lead to numerous developments. Unfortunately, the description of Bell inequalities is subject to several degeneracies, which make any exchange of information about them unnecessarily hard. Here, we analyze these degeneracies and propose a decomposition for Bell-like inequalities based on a set of reference expressions which is not affected by them. These reference expressions set a common ground for comparing Bell inequalities. We provide algorithms based on finite group theory to compute this decomposition. Implementing these algorithms allows us to set up a compendium of reference Bell-like inequalities, available online at http://www.faacets.com . This website constitutes a platform where registered Bell-like inequalities can be explored, new inequalities can be compared to previously-known ones and relevant information on Bell inequalities can be added in a collaborative manner.
INTRODUCTIONThe years 1990s started with the seminal paper presenting the Ekert'91 protocol [2], relating quantum nonlocality to secure communication. This changed the world as far as quantum nonlocality is concerned; the study of Bell inequalities became respectable. So far not much was known beyond the famous CHSH inequality [3]. Here it is noteworthy to mention that Bell's original inequality published in 1964 [1] is not a Bell inequality in the modern sense, because it relies on the additional assumption of perfect anti-correlation when both sides perform the same measurement. In particular, little was known when the parties perform measurements with more than two possible outcomes. Kaszlikowski and co-workers performed numerical searches for experimental scenarios more resistant to noise [4]; this effort led Dan Collins, then at Geneva University, and colleagues to find the family of inequalities behind Kaszlikowski et al. finding, today known as the CGLMP inequalities [5]. Meanwhile, Pitowski and Svozil, building on their understanding that the set of local correlations constitues a polytope, could find all the inequalities corresponding to the facets of two scenarios of interest [6]. In a subsequent work, Sliwa [8] and grouped the results of Pitowski and Svozil into families of inequalities equivalent under relabelings. In particular Sliwa found all the families corresponding to the scenario with 3 parties and binary inputs and outcomes, while Collins-Gisin found, among others, the family known as I nnmm . Avis, Imai, Ito and Sakasi found many more Bell inequalities using specialized cut-polytopes [9]. And so the field expanded very significantly, though it would still be nice to have more families of inequalities valid for arbitrary number of parties, measurement settings and outcomes [10][11][12][13][14]. Also, experiments on Bell's inequalities went out of the lab and entered applied physics [15][16][17][18].Another trend that started was the use of Bell-like inequal...