arXiv subjects

Maxime Meyer

Publications and source records attributed to Maxime Meyer.

13 recordsLinked to original sources

arXiv2026

How Many Different Outputs Can a Transformer Generate?

We study how we can leverage only a handful of characteristics of a transformer's architecture to closely predict the number of different sequences it can output, both qualitatively and quantitatively. We provide an upper bound depending on the length of the prompt, which we show empirically to be tight up to a factor less than 10, across architectures and model sizes. Our analysis also provides a theoretical explanation for previously observed empirical failures of transformers on simple sequence tasks, such as copying and cramming. Formally, we prove that (i) the maximal length of accessible sequences (those that the transformer can output for some prompt) grows linearly with the prompt length, (ii) beyond a critical threshold, the proportion of accessible sequences decays exponentially with sequence length, and (iii) the linear coefficient relating prompt length to accessible sequence length admits a theoretical upper bound. Notably, these results hold even with unbounded context and computation time.

cs.LG↗

arXiv2026

Online Learning-to-Defer with Varying Experts

Learning-to-Defer (L2D) methods route each query either to a predictive model or to external experts. Real-world deployments require handling streaming data, changing expert availability, shifting expert reliability, and feedback observed only for the selected action. We introduce an online multiclass L2D algorithm that combines queried-action bandit feedback with a dynamically varying pool of experts. Let $N=n+n_e$, let $B$ bound the Frobenius norm of the linear score matrix, and let $\rho$ bound the augmented input norm. Assuming linear calibration and zero surrogate minimizability gap for the projected comparator class, our method achieves expected true-deferral regret $O((BN^{3/2}\rho+1)T^{2/3})$, improving to $O(BN^{3/2}\rho\sqrt T+B^2N^3\rho^2)$ under a concentrated-score condition. The analysis combines an online $\mathcal H$-consistency transfer bound with projected online convex optimization. Experiments on synthetic and real-world datasets demonstrate selective routing under varying expert availability and reliability.

stat.ML↗

arXiv2026

Distilling Human-Aligned Privacy Sensitivity Assessment from Large Language Models

Accurate privacy evaluation of textual data remains a critical challenge in privacy-preserving natural language processing. Recent work has shown that large language models (LLMs) can serve as reliable privacy evaluators, achieving strong agreement with human judgments; however, their computational cost and impracticality for processing sensitive data at scale limit real-world deployment. We address this gap by distilling the privacy assessment capabilities of Mistral Large 3 (675B) into lightweight encoder models with as few as 150M parameters. Leveraging a large-scale dataset of privacy-annotated texts spanning 10 diverse domains, we train efficient classifiers that preserve strong agreement with human annotations while dramatically reducing computational requirements. We validate our approach on human-annotated test data and demonstrate its practical utility as an evaluation metric for de-identification systems.

cs.CL↗

arXiv2026

Adaptive Text Anonymization: Learning Privacy-Utility Trade-offs via Prompt Optimization

Anonymizing textual documents is a highly context-sensitive problem: the appropriate balance between privacy protection and utility preservation varies with the data domain, privacy objectives, and downstream application. However, existing anonymization methods rely on static, manually designed strategies that lack the flexibility to adjust to diverse requirements and often fail to generalize across domains. We introduce adaptive text anonymization, a new task formulation in which anonymization strategies are automatically adapted to specific privacy-utility requirements. We propose a framework for task-specific prompt optimization that automatically constructs anonymization instructions for language models, enabling adaptation to different privacy goals, domains, and downstream usage patterns. To evaluate our approach, we present a benchmark spanning five datasets with diverse domains, privacy constraints, and utility objectives. Across all evaluated settings, our framework consistently achieves a better privacy-utility trade-off than existing baselines, while remaining computationally efficient and effective on open-source language models, with performance comparable to larger closed-source models. Additionally, we show that our method can discover novel anonymization strategies that explore different points along the privacy-utility trade-off frontier.

cs.CL↗

arXiv2025

Memory Limitations of Prompt Tuning in Transformers

Despite the empirical success of prompt tuning in adapting pretrained language models to new tasks, theoretical analyses of its capabilities remain limited. Existing theoretical work primarily addresses universal approximation properties, demonstrating results comparable to standard weight tuning. In this paper, we explore a different aspect of the theory of transformers: the memorization capability of prompt tuning. We provide two principal theoretical contributions. First, we prove that the amount of information memorized by a transformer cannot scale faster than linearly with the prompt length. Second, and more importantly, we present the first formal proof of a phenomenon empirically observed in large language models: performance degradation in transformers with extended contexts. We rigorously demonstrate that transformers inherently have limited memory, constraining the amount of information they can retain, regardless of the context size. This finding offers a fundamental understanding of the intrinsic limitations of transformer architectures, particularly their ability to handle long sequences.

cs.LG↗

arXiv2025

Tau-Eval: A Unified Evaluation Framework for Useful and Private Text Anonymization

Text anonymization is the process of removing or obfuscating information from textual data to protect the privacy of individuals. This process inherently involves a complex trade-off between privacy protection and information preservation, where stringent anonymization methods can significantly impact the text's utility for downstream applications. Evaluating the effectiveness of text anonymization proves challenging from both privacy and utility perspectives, as there is no universal benchmark that can comprehensively assess anonymization techniques across diverse, and sometimes contradictory contexts. We present Tau-Eval, an open-source framework for benchmarking text anonymization methods through the lens of privacy and utility task sensitivity. A Python library, code, documentation and tutorials are publicly available.

cs.CL↗

arXiv2025

Online Learning of Pure States is as Hard as Mixed States

Quantum state tomography, the task of learning an unknown quantum state, is a fundamental problem in quantum information. In standard settings, the complexity of this problem depends significantly on the type of quantum state that one is trying to learn, with pure states being substantially easier to learn than general mixed states. A natural question is whether this separation holds for any quantum state learning setting. In this work, we consider the online learning framework and prove the surprising result that learning pure states in this setting is as hard as learning mixed states. More specifically, we show that both classes share almost the same sequential fat-shattering dimension, leading to identical regret scaling. We also generalize previous results on full quantum state tomography in the online setting to (i) the $\epsilon$-realizable setting and (ii) learning the density matrix only partially, using smoothed analysis.

quant-ph↗

arXiv2024

TAROT: Task-Oriented Authorship Obfuscation Using Policy Optimization Methods

Authorship obfuscation aims to disguise the identity of an author within a text by altering the writing style, vocabulary, syntax, and other linguistic features associated with the text author. This alteration needs to balance privacy and utility. While strong obfuscation techniques can effectively hide the author's identity, they often degrade the quality and usefulness of the text for its intended purpose. Conversely, maintaining high utility tends to provide insufficient privacy, making it easier for an adversary to de-anonymize the author. Thus, achieving an optimal trade-off between these two conflicting objectives is crucial. In this paper, we propose TAROT: Task-Oriented Authorship Obfuscation Using Policy Optimization, a new unsupervised authorship obfuscation method whose goal is to optimize the privacy-utility trade-off by regenerating the entire text considering its downstream utility. Our approach leverages policy optimization as a fine-tuning paradigm over small language models in order to rewrite texts by preserving author identity and downstream task utility. We show that our approach largely reduces the accuracy of attackers while preserving utility. We make our code and models publicly available.

cs.CL↗

arXiv2023

Targeted Attacks: Redefining Spear Phishing and Business Email Compromise

In today's digital world, cybercrime is responsible for significant damage to organizations, including financial losses, operational disruptions, or intellectual property theft. Cyberattacks often start with an email, the major means of corporate communication. Some rare, severely damaging email threats - known as spear phishing or Business Email Compromise - have emerged. However, the literature disagrees on their definition, impeding security vendors and researchers from mitigating targeted attacks. Therefore, we introduce targeted attacks. We describe targeted-attack-detection techniques as well as social-engineering methods used by fraudsters. Additionally, we present text-based attacks - with textual content as malicious payload - and compare non-targeted and targeted variants.

cs.CR↗

arXiv2022

NL-Augmenter: A Framework for Task-Sensitive Natural Language Augmentation

Data augmentation is an important component in the robustness evaluation of models in natural language processing (NLP) and in enhancing the diversity of the data they are trained on. In this paper, we present NL-Augmenter, a new participatory Python-based natural language augmentation framework which supports the creation of both transformations (modifications to the data) and filters (data splits according to specific features). We describe the framework and an initial set of 117 transformations and 23 filters for a variety of natural language tasks. We demonstrate the efficacy of NL-Augmenter by using several of its transformations to analyze the robustness of popular natural language models. The infrastructure, datacards and robustness analysis results are available publicly on the NL-Augmenter repository (https://github.com/GEM-benchmark/NL-Augmenter).

cs.CL↗

arXiv2021

Text Data Augmentation: Towards better detection of spear-phishing emails

Text data augmentation, i.e., the creation of new textual data from an existing text, is challenging. Indeed, augmentation transformations should take into account language complexity while being relevant to the target Natural Language Processing (NLP) task (e.g., Machine Translation, Text Classification). Initially motivated by an application of Business Email Compromise (BEC) detection, we propose a corpus and task agnostic augmentation framework used as a service to augment English texts within our company. Our proposal combines different methods, utilizing BERT language model, multi-step back-translation and heuristics. We show that our augmentation framework improves performances on several text classification tasks using publicly available models and corpora as well as on a BEC detection task. We also provide a comprehensive argumentation about the limitations of our augmentation framework.

cs.CL↗

arXiv2019

An Overview of GSMA's M2M Remote Provisioning Specification

M2M devices are ubiquitous, and there is a growing tendency to connect such devices to mobile networks. Network operators are investigating new solutions to lower their costs and to address usability issues. Embedded SIM cards with remote provisioning capability are one of the most promising solutions. GSMA, the leading consortium on mobile network standards, has proposed a specification for such an embedded SIM card, called eUICC. The specification describes eUICC architecture and a remote provisioning mechanism. Embodiments of this specification have the potential to disrupt the telecommunications market: eUICCs will be shipped to device manufacturers and then remotely provisioned with a subscription, whereas (currently) SIMs must be provisioned prior to shipping. In this article, we present a comprehensive overview of GSMA's specification and its motivation. In particular, we describe the technology and the protocols involved in remote provisioning.

cs.NI↗

arXiv2018

Exploiting re-voting in the Helios election system

Election systems must ensure that representatives are chosen by voters. Moreover, each voter should have equal influence. Traditionally, this has been achieved by permitting voters to cast at most one ballot. More recently, this has been achieved by tallying the last ballot cast by each voter. We show this is not achieved by the Helios election system, because an adversary can cause a ballot other than a voter's last to be tallied. Moreover, we show how the adversary can choose the contents of such a ballot, thus the adversary can unduly influence the selection of representatives.

cs.CR↗