Measurement of CP asymmetries in $\lambda^0_b \to pk^-$ and $\lambda^0_b \to p \pi^-$ decays at LHCb

The LHCb experiment has been designed to perform precision measurements in the flavour physics sector at the Large Hadron Collider (LHC) located at CERN. After the recent observation of CP violation in the decay of the Bs0 meson to a charged pion-kaon pair at LHCb, it is interesting to see whether the same quark-level transition in Λ0b baryon decays gives rise to large CP-violating effects. Such decay processes involve both tree and penguin Feynman diagrams and could be sensitive probes for physics beyond the Standard Model. The measurement of the CP-violating observable defined as ∆ACP = ACP(Λ0b → pK−)−ACP(Λ0b →pπ−),where ACP(Λ0b →pK−) and ACP(Λ0b →pπ−) are the direct CP asymmetries in Λ0b → pK− and Λ0b → pπ− decays, is presented for the first time using LHCb data. The procedure followed to optimize the event selection, to calibrate particle identification, to parametrise the various components of the invariant mass spectra, and to compute corrections due to the production asymmetry of the initial state and the detection asymmetries of the final states, is discussed in detail. Using the full 2011 and 2012 data sets of pp collisions collected with the LHCb detector, corresponding to an integrated luminosity of about 3 fb−1, the value ∆ACP = (0.8 ± 2.1 ± 0.2)% is obtained. The first uncertainty is statistical and the second corresponds to one of the dominant systematic effects. As the result is compatible with zero, no evidence of CP violation is found. This is the most precise measurement of CP violation in the decays of baryons containing the b quark to date. Once the analysis will be completed with an exhaustive study of systematic uncertainties, the results will be published by the LHCb Collaboration.

Asymmetries in Adversarial Settings

Even without formal guarantees of their effectiveness, adversarial attacks against Machine Learning models frequently fool new defenses. We identify six key asymmetries that contribute to this phenomenon and formulate four guidelines to build future-proof defenses by preventing such asymmetries. We also prove that attacking a classifier is NP-complete, while defending from such attacks is Sigma_2^P-complete. We then introduce Counter-Attack (CA), an asymmetry-free metadefense that determines whether a model is robust on a given input by estimating its distance from the decision boundary. Under specific assumptions CA can provide theoretical detection guarantees. Additionally, we prove that while CA is NP-complete, fooling CA is Sigma_2^P-complete. Even when using heuristic relaxations, we show that our method can reliably identify non-robust points. As part of our experimental evaluation, we introduce UG100, a new dataset obtained by applying a provably optimal attack to six limited-scale networks (three for MNIST and three for CIFAR10), each trained in three different manners.