Yannick Chevalier

Fanny Jourdan, Yannick Chevalier, Cécile Favre

Large Language Models (LLMs) are increasingly leveraged for translation tasks but often fall short when translating inclusive language -- such as texts containing the singular 'they' pronoun or otherwise reflecting fair linguistic protocols. Because these challenges span both computational and societal domains, it is imperative to critically evaluate how well LLMs handle inclusive translation with a well-founded framework. This paper presents FairTranslate, a novel, fully human-annotated dataset designed to evaluate non-binary gender biases in machine translation systems from English to French. FairTranslate includes 2418 English-French sentence pairs related to occupations, annotated with rich metadata such as the stereotypical alignment of the occupation, grammatical gender indicator ambiguity, and the ground-truth gender label (male, female, or inclusive). We evaluate four leading LLMs (Gemma2-2B, Mistral-7B, Llama3.1-8B, Llama3.3-70B) on this dataset under different prompting procedures. Our results reveal substantial biases in gender representation across LLMs, highlighting persistent challenges in achieving equitable outcomes in machine translation. These findings underscore the need for focused strategies and interventions aimed at ensuring fair and inclusive language usage in LLM-based translation systems. We make the FairTranslate dataset publicly available on Hugging Face, and disclose the code for all experiments on GitHub.

Yannick Chevalier, Michaël Rusinowitch

Cryptographic protocols are often specified by narrations, i.e., finite sequences of message exchanges that show the intended execution of the protocol. Another use of narrations is to describe attacks. We propose in this paper to compile, when possible, attack describing narrations into a set of tests that honest participants can perform to exclude these executions. These tests can be implemented in monitors to protect existing implementations from rogue behaviour.

Tigran Avanesov, Yannick Chevalier, Michaël Rusinowitch et al.

The problem of finding a mediator to compose secured services has been reduced in our former work to the problem of solving deducibility constraints similar to those employed for cryptographic protocol analysis. We extend in this paper the mediator synthesis procedure by a construction for expressing that some data is not accessible to the mediator. Then we give a decision procedure for verifying that a mediator satisfying this non-disclosure policy can be effectively synthesized. This procedure has been implemented in CL-AtSe, our protocol analysis tool. The procedure extends constraint solving for cryptographic protocol analysis in a significative way as it is able to handle negative deducibility constraints without restriction. In particular it applies to all subterm convergent theories and therefore covers several interesting theories in formal security analysis including encryption, hashing, signature and pairing.