Imagen de Portada

SHIELD: Classifier-Guided Prompting for Robust and Safer LVLMs

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Ren, Juan, Dras, Mark, Naseem, Usman
Formato:	Preprint
Publicado:	2025
Materias:	Computation and Language
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Seeing the Threat: Vulnerabilities in Vision-Language Models to Adversarial Attack
por: Ren, Juan, et al.
Publicado: (2025)

Agentic Moderation: Multi-Agent Design for Safer Vision-Language Models
por: Ren, Juan, et al.
Publicado: (2025)

Should LLM Safety Be More Than Refusing Harmful Instructions?
por: Maskey, Utsav, et al.
Publicado: (2025)

Steering Over-refusals Towards Safety in Retrieval Augmented Generation
por: Maskey, Utsav, et al.
Publicado: (2025)

Over-Refusal and Representation Subspaces: A Mechanistic Analysis of Task-Conditioned Refusal in Aligned LLMs
por: Maskey, Utsav, et al.
Publicado: (2026)

Steering Towards Fairness: Mitigating Political Bias in LLMs
por: Nadeem, Afrozah, et al.
Publicado: (2025)

Fairness Evaluation and Inference Level Mitigation in LLMs
por: Nadeem, Afrozah, et al.
Publicado: (2025)

Framing Political Bias in Multilingual LLMs Across Pakistani Languages
por: Nadeem, Afrozah, et al.
Publicado: (2025)

We Think, Therefore We Align LLMs to Helpful, Harmless and Honest Before They Go Wrong
por: Kashyap, Gautam Siddharth, et al.
Publicado: (2025)

Too Helpful, Too Harmless, Too Honest or Just Right?
por: Kashyap, Gautam Siddharth, et al.
Publicado: (2025)

AlignCultura: Towards Culturally Aligned Large Language Models?
por: Kashyap, Gautam Siddharth, et al.
Publicado: (2026)

When the Model Said 'No Comment', We Knew Helpfulness Was Dead, Honesty Was Alive, and Safety Was Terrified
por: Kashyap, Gautam Siddharth, et al.
Publicado: (2026)

Beyond the Black Box: Demystifying Multi-Turn LLM Reasoning with VISTA
por: Zhang, Yiran, et al.
Publicado: (2025)

CogMem: A Cognitive Memory Architecture for Sustained Multi-Turn Reasoning in Large Language Models
por: Zhang, Yiran, et al.
Publicado: (2025)

SafeConstellations: Mitigating Over-Refusals in LLMs Through Task-Aware Representation Steering
por: Maskey, Utsav, et al.
Publicado: (2025)

VITAL: A New Dataset for Benchmarking Pluralistic Alignment in Healthcare
por: Shetty, Anudeex, et al.
Publicado: (2025)

DUAL-Bench: Measuring Over-Refusal and Robustness in Vision-Language Models
por: Ren, Kaixuan, et al.
Publicado: (2025)

A Survey on Progress in LLM Alignment from the Perspective of Reward Design
por: Ji, Miaomiao, et al.
Publicado: (2025)

VaxGuard: A Multi-Generator, Multi-Type, and Multi-Role Dataset for Detecting LLM-Generated Vaccine Misinformation
por: Ahmad, Syed Talal, et al.
Publicado: (2025)

Mechanistic Interpretability for Large Language Model Alignment: Progress, Challenges, and Future Directions
por: Naseem, Usman
Publicado: (2026)

Myanmar XNLI: Building a Dataset and Exploring Low-resource Approaches to Natural Language Inference with Myanmar
por: Htet, Aung Kyaw, et al.
Publicado: (2025)

Benchmarking Large Language Models for Cryptanalysis and Side-Channel Vulnerabilities
por: Maskey, Utsav, et al.
Publicado: (2025)

Enhancing ESG Impact Type Identification through Early Fusion and Multilingual Models
por: Veeramani, Hariram, et al.
Publicado: (2024)

PersoPilot: An Adaptive AI-Copilot for Transparent Contextualized Persona Classification and Personalized Response Generation
por: Afzoon, Saleh, et al.
Publicado: (2026)

Safer in Translation? Presupposition Robustness in Indic Languages
por: Palnitkar, Aadi, et al.
Publicado: (2025)

Reversal of Thought: Enhancing Large Language Models with Preference-Guided Reverse Reasoning Warm-up
por: Yuan, Jiahao, et al.
Publicado: (2024)

ExtremeAIGC: Benchmarking LMM Vulnerability to AI-Generated Extremist Content
por: Chandna, Bhavik, et al.
Publicado: (2025)

When Prompts Override Vision: Prompt-Induced Hallucinations in LVLMs
por: Khayatan, Pegah, et al.
Publicado: (2026)

TurnBench-MS: A Benchmark for Evaluating Multi-Turn, Multi-Step Reasoning in Large Language Models
por: Zhang, Yiran, et al.
Publicado: (2025)

XGUARD: A Graded Benchmark for Evaluating Safety Failures of Large Language Models on Extremist Content
por: Abishethvarman, Vadivel, et al.
Publicado: (2025)

PersoBench: Benchmarking Personalized Response Generation in Large Language Models
por: Afzoon, Saleh, et al.
Publicado: (2024)

Evaluating Hierarchical Clinical Document Classification Using Reasoning-Based LLMs
por: Mustafa, Akram, et al.
Publicado: (2025)

Can Reasoning LLMs Enhance Clinical Document Classification?
por: Mustafa, Akram, et al.
Publicado: (2025)

Graded Suspiciousness of Adversarial Texts to Human
por: Tonni, Shakila Mahjabin, et al.
Publicado: (2024)

Safer or Luckier? LLMs as Safety Evaluators Are Not Robust to Artifacts
por: Chen, Hongyu, et al.
Publicado: (2025)

MRG-R1: Reinforcement Learning for Clinically Aligned Medical Report Generation
por: Wang, Pengyu, et al.
Publicado: (2025)

Bias Beyond Borders: Political Ideology Evaluation and Steering in Multilingual LLMs
por: Nadeem, Afrozah, et al.
Publicado: (2026)

Do LVLMs Know What They Know? A Systematic Study of Knowledge Boundary Perception in LVLMs
por: Ding, Zhikai, et al.
Publicado: (2025)

PersoDPO: Scalable Preference Optimization for Instruction-Adherent, Persona-Grounded Dialogue via Multi-LLM Evaluation
por: Afzoon, Saleh, et al.
Publicado: (2026)

LVMed-R2: Perception and Reflection-driven Complex Reasoning for Medical Report Generation
por: Wang, Hao, et al.
Publicado: (2025)