Image de couverture de livre

LLMs Lost in Translation: M-ALERT uncovers Cross-Linguistic Safety Inconsistencies

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Friedrich, Felix, Tedeschi, Simone, Schramowski, Patrick, Brack, Manuel, Navigli, Roberto, Nguyen, Huu, Li, Bo, Kersting, Kristian
Format:	Preprint
Publié:	2024
Sujets:	Computation and Language
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

ALERT: A Comprehensive Benchmark for Assessing Large Language Models' Safety through Red Teaming
par: Tedeschi, Simone, et autres
Publié: (2024)

SCAR: Sparse Conditioned Autoencoders for Concept Detection and Steering in LLMs
par: Härle, Ruben, et autres
Publié: (2024)

LlavaGuard: An Open VLM-based Framework for Safeguarding Vision Datasets and Models
par: Helff, Lukas, et autres
Publié: (2024)

Beyond Overcorrection: Evaluating Diversity in T2I Models with DivBench
par: Friedrich, Felix, et autres
Publié: (2025)

T-FREE: Subword Tokenizer-Free Generative LLMs via Sparse Representations for Memory-Efficient Embeddings
par: Deiseroth, Björn, et autres
Publié: (2024)

Exploiting Cultural Biases via Homoglyphs in Text-to-Image Synthesis
par: Struppek, Lukas, et autres
Publié: (2022)

Does CLIP Know My Face?
par: Hintersdorf, Dominik, et autres
Publié: (2022)

LIME: Making LLM Data More Efficient with Linguistic Metadata Embeddings
par: Sztwiertnia, Sebastian, et autres
Publié: (2025)

Core Tokensets for Data-efficient Sequential Training of Transformers
par: Paul, Subarnaduti, et autres
Publié: (2024)

How to Train your Text-to-Image Model: Evaluating Design Choices for Synthetic Training Captions
par: Brack, Manuel, et autres
Publié: (2025)

Measuring and Guiding Monosemanticity
par: Härle, Ruben, et autres
Publié: (2025)

LEDITS++: Limitless Image Editing using Text-to-Image Models
par: Brack, Manuel, et autres
Publié: (2023)

Multilingual Text-to-Image Generation Magnifies Gender Stereotypes and Prompt Engineering May Not Help You
par: Friedrich, Felix, et autres
Publié: (2024)

A Typology for Exploring the Mitigation of Shortcut Behavior
par: Friedrich, Felix, et autres
Publié: (2022)

AtMan: Understanding Transformer Predictions Through Memory Efficient Attention Manipulation
par: Deiseroth, Björn, et autres
Publié: (2023)

DeiSAM: Segment Anything with Deictic Prompting
par: Shindo, Hikaru, et autres
Publié: (2024)

No Safe Dose: How Training Data Drives Unsafe Image Generation
par: Friedrich, Felix, et autres
Publié: (2026)

ActivationReasoning: Logical Reasoning in Latent Activation Spaces
par: Helff, Lukas, et autres
Publié: (2025)

CHRONOBERG: Capturing Language Evolution and Temporal Awareness in Foundation Models
par: Hegde, Niharika, et autres
Publié: (2025)

Learning by Self-Explaining
par: Stammer, Wolfgang, et autres
Publié: (2023)

LLMs Gaming Verifiers: RLVR can Lead to Reward Hacking
par: Helff, Lukas, et autres
Publié: (2026)

SocialGrid: A Benchmark for Planning and Social Reasoning in Embodied Multi-Agent Systems
par: Shindo, Hikaru, et autres
Publié: (2026)

Adaptive Rational Activations to Boost Deep Reinforcement Learning
par: Delfosse, Quentin, et autres
Publié: (2021)

ART: Adaptive Relation Tuning for Generalized Relation Prediction
par: Sudhakaran, Gopika, et autres
Publié: (2025)

The Cake that is Intelligence and Who Gets to Bake it: An AI Analogy and its Implications for Participation
par: Mundt, Martin, et autres
Publié: (2025)

Right Answer, Wrong Score: Uncovering the Inconsistencies of LLM Evaluation in Multiple-Choice Question Answering
par: Molfese, Francesco Maria, et autres
Publié: (2025)

Divergent Token Metrics: Measuring degradation to prune away LLM components -- and optimize quantization
par: Deiseroth, Björn, et autres
Publié: (2023)

Truth or Mirage? Towards End-to-End Factuality Evaluation with LLM-Oasis
par: Scirè, Alessandro, et autres
Publié: (2024)

SLR: Automated Synthesis for Scalable Logical Reasoning
par: Helff, Lukas, et autres
Publié: (2025)

Has Machine Translation Evaluation Achieved Human Parity? The Human Reference and the Limits of Progress
par: Proietti, Lorenzo, et autres
Publié: (2025)

Learning to Intervene on Concept Bottlenecks
par: Steinmann, David, et autres
Publié: (2023)

EmoNet-Voice: A Fine-Grained, Expert-Verified Benchmark for Speech Emotion Detection
par: Schuhmann, Christoph, et autres
Publié: (2025)

Estimating Machine Translation Difficulty
par: Proietti, Lorenzo, et autres
Publié: (2025)

Lost in Multilinguality: Dissecting Cross-lingual Factual Inconsistency in Transformer Language Models
par: Wang, Mingyang, et autres
Publié: (2025)

Process Reward Models Meet Planning: Generating Precise and Scalable Datasets for Step-Level Rewards
par: Pisano, Raffaele, et autres
Publié: (2026)

EmoNet-Face: An Expert-Annotated Benchmark for Synthetic Emotion Recognition
par: Schuhmann, Christoph, et autres
Publié: (2025)

Making deep neural networks right for the right scientific reasons by interacting with their explanations
par: Schramowski, Patrick, et autres
Publié: (2020)

Guardians of the Machine Translation Meta-Evaluation: Sentinel Metrics Fall In!
par: Perrella, Stefano, et autres
Publié: (2024)

Multipole Semantic Attention: A Fast Approximation of Softmax Attention for Pretraining
par: Mitchell, Rupert, et autres
Publié: (2025)

ZEBRA: Zero-Shot Example-Based Retrieval Augmentation for Commonsense Question Answering
par: Molfese, Francesco Maria, et autres
Publié: (2024)