Imagen de Portada

LLM-VA: Resolving the Jailbreak-Overrefusal Trade-off via Vector Alignment

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Zhang, Haonan, Wang, Dongxia, Liu, Yi, Chen, Kexin, Wang, Wenhai
Formato:	Preprint
Publicado:	2026
Materias:	Machine Learning Artificial Intelligence
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Characterizing and Evaluating the Reliability of LLMs against Jailbreak Attacks
por: Chen, Kexin, et al.
Publicado: (2024)

Sticking to the Mean: Detecting Sticky Tokens in Text Embedding Models
por: Chen, Kexin, et al.
Publicado: (2025)

KG4RecEval: Does Knowledge Graph Really Matter for Recommender Systems?
por: Zhang, Haonan, et al.
Publicado: (2024)

Understanding and Mitigating Overrefusal in LLMs from an Unveiling Perspective of Safety Decision Boundary
por: Pan, Licheng, et al.
Publicado: (2025)

ORFuzz: Fuzzing the "Other Side" of LLM Safety -- Testing Over-Refusal
por: Zhang, Haonan, et al.
Publicado: (2025)

Reasoned Safety Alignment: Ensuring Jailbreak Defense via Answer-Then-Check
por: Cao, Chentao, et al.
Publicado: (2025)

Deactivating Refusal Triggers: Understanding and Mitigating Overrefusal in Safety Alignment
por: Xue, Zhiyu, et al.
Publicado: (2026)

Functional Homotopy: Smoothing Discrete Optimization via Continuous Parameters for LLM Jailbreak Attacks
por: Wang, Zi, et al.
Publicado: (2024)

Navigating the Alignment-Calibration Trade-off: A Pareto-Superior Frontier via Model Merging
por: Hu, Tiancheng, et al.
Publicado: (2025)

Aggregation Alignment for Federated Learning with Mixture-of-Experts under Data Heterogeneity
por: Fang, Zihan, et al.
Publicado: (2026)

Uncovering Logit Suppression Vulnerabilities in LLM Safety Alignment
por: Li, Yuxi, et al.
Publicado: (2024)

Adversarial Preference Learning for Robust LLM Alignment
por: Wang, Yuanfu, et al.
Publicado: (2025)

Reasoning-targeted Jailbreak Attacks on Large Reasoning Models via Semantic Triggers and Psychological Framing
por: Wang, Zehao, et al.
Publicado: (2026)

VP-VAE: Rethinking Vector Quantization via Adaptive Vector Perturbation
por: Zhai, Linwei, et al.
Publicado: (2026)

Learning Grouped Lattice Vector Quantizers for Low-Bit LLM Compression
por: Zhang, Xi, et al.
Publicado: (2025)

Enhancing few-shot time series forecasting with LLM-guided diffusion
por: Shi, Haonan, et al.
Publicado: (2026)

Adaptive Helpfulness-Harmlessness Alignment with Preference Vectors
por: Liang, Ren-Wei, et al.
Publicado: (2025)

Immune: Improving Safety Against Jailbreaks in Multi-modal LLMs via Inference-Time Alignment
por: Ghosal, Soumya Suvra, et al.
Publicado: (2024)

Quantile Advantage Estimation: Stabilizing RLVR for LLM Reasoning
por: Wu, Junkang, et al.
Publicado: (2025)

Enhancing LLM Steering through Sparse Autoencoder-Based Vector Refinement
por: Wang, Anyi, et al.
Publicado: (2025)

Safeguarding LLM Fine-tuning via Push-Pull Distributional Alignment
por: Wang, Haozhong, et al.
Publicado: (2026)

Light Alignment Improves LLM Safety via Model Self-Reflection with a Single Neuron
por: Shen, Sicheng, et al.
Publicado: (2026)

Automating Deception: Scalable Multi-Turn LLM Jailbreaks
por: Kumarappan, Adarsh, et al.
Publicado: (2025)

From Entanglement to Alignment: Representation Space Decomposition for Unsupervised Time Series Domain Adaptation
por: Cai, Rongyao, et al.
Publicado: (2025)

MAnchors: Memorization-Based Acceleration of Anchors via Rule Reuse and Transformation
por: Yu, Haonan, et al.
Publicado: (2025)

Conformal Feedback Alignment: Quantifying Answer-Level Reliability for Robust LLM Alignment
por: Chen, Tiejin, et al.
Publicado: (2026)

Metis: Learning to Jailbreak LLMs via Self-Evolving Metacognitive Policy Optimization
por: Zhou, Huilin, et al.
Publicado: (2026)

ALERT: Zero-shot LLM Jailbreak Detection via Internal Discrepancy Amplification
por: Lin, Xiao, et al.
Publicado: (2026)

Trading-off Accuracy and Communication Cost in Federated Learning
por: Villani, Mattia Jacopo, et al.
Publicado: (2025)

The Art of the Jailbreak: Formulating Jailbreak Attacks for LLM Security Beyond Binary Scoring
por: Hossain, Ismail, et al.
Publicado: (2026)

Alignment Dynamics in LLM Fine-Tuning
por: Huang, Yuhan, et al.
Publicado: (2026)

Beyond Myopia: Learning from Positive and Unlabeled Data through Holistic Predictive Trends
por: Wang, Xinrui, et al.
Publicado: (2023)

SmoothLLM: Defending Large Language Models Against Jailbreaking Attacks
por: Robey, Alexander, et al.
Publicado: (2023)

Rounding-Guided Backdoor Injection in Deep Learning Model Quantization
por: Chen, Xiangxiang, et al.
Publicado: (2025)

Generalist++: A Meta-learning Framework for Mitigating Trade-off in Adversarial Training
por: Wang, Yisen, et al.
Publicado: (2025)

Divide and Learn: Multi-Objective Combinatorial Optimization at Scale
por: Singh, Esha, et al.
Publicado: (2026)

Auditing and Generating Synthetic Data with Controllable Trust Trade-offs
por: Belgodere, Brian, et al.
Publicado: (2023)

Learning Heterogeneous Performance-Fairness Trade-offs in Federated Learning
por: Ye, Rongguang, et al.
Publicado: (2025)

ECLIPTICA -- A Framework for Switchable LLM Alignment via CITA - Contrastive Instruction-Tuned Alignment
por: Wanaskar, Kapil, et al.
Publicado: (2026)

MarkTune: Improving the Quality-Detectability Trade-off in Open-Weight LLM Watermarking
por: Zhao, Yizhou, et al.
Publicado: (2025)