In the scope of discrete finite-state models of interacting components, we present a novel algorithm for identifying sets of local states of components whose activity is necessary for the reachability of a given local state. If all the local states from such a set are disabled in the model, the concerned reachability is impossible. Those sets are referred to as cut sets and are computed from a particular abstract causality structure, so-called Graph of Local Causality, inspired from previous work and generalised here to finite automata networks. The extracted sets of local states form an under-approximation of the complete minimal cut sets of the dynamics: there may exist smaller or additional cut sets for the given reachability. Applied to qualitative models of biological systems, such cut sets provide potential therapeutic targets that are proven to prevent molecules of interest to become active, up to the correctness of the model. Our new method makes tractable the formal analysis of very large scale networks, as illustrated by the computation of cut sets within a Boolean model of biological pathways interactions gathering more than 9000 components.

We address the problem of Reliable Broadcast in asynchronous message-passing systems with $n$ nodes, of which up to $t$ are malicious (faulty), in addition to a message adversary that can drop some of the messages sent by correct (non-faulty) nodes. We present a Message-Adversary-Tolerant Byzantine Reliable Broadcast (MBRB) algorithm that communicates ${\cal O}(|m|+n\kappa)$ bits per node, where $|m|$ represents the length of the application message and $\kappa=\Omega(\log n)$ is a security parameter. This communication complexity is optimal up to the parameter $\kappa$. This significantly improves upon the state-of-the-art MBRB solution (Albouy, Frey, Raynal, and Ta\"iani, TCS 2023), which incurs communication of ${\cal O}(n|m|+n^2\kappa)$ bits per node. Our solution sends at most $4n^2$ messages overall, which is asymptotically optimal. Reduced communication is achieved by employing coding techniques that replace the need for all nodes to (re-)broadcast the entire application message $m$. Instead, nodes forward authenticated fragments of the encoding of $m$ using an erasure-correcting code. Under the cryptographic assumptions of threshold signatures and vector commitments, and assuming $n > 3t+2d$, where the adversary drops at most $d$ messages per broadcast, our algorithm allows at least $\ell = n - t - (1 + \epsilon)d$ (for any arbitrarily low $\epsilon> 0$) correct nodes to reconstruct $m$, despite missing fragments caused by the malicious nodes and the message adversary.

With the introduction of (large) language models, there has been significant concern about the unintended bias such models may inherit from their training data. A number of studies have shown that such models propagate gender stereotypes, as well as geographical and racial bias, among other biases. While existing works tackle this issue by preprocessing data and debiasing embeddings, the proposed methods require a lot of computational resources and annotation effort while being limited to certain types of biases. To address these issues, we introduce REFINE-LM, a debiasing method that uses reinforcement learning to handle different types of biases without any fine-tuning. By training a simple model on top of the word probability distribution of a LM, our bias agnostic reinforcement learning method enables model debiasing without human annotations or significant computational resources. Experiments conducted on a wide range of models, including several LMs, show that our method (i) significantly reduces stereotypical biases while preserving LMs performance; (ii) is applicable to different types of biases, generalizing across contexts such as gender, ethnicity, religion, and nationality-based biases; and (iii) it is not expensive to train.