Imagen de Portada

Guard Vector: Beyond English LLM Guardrails with Task-Vector Composition and Streaming-Aware Prefix SFT

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Lee, Wonhyuk, Kim, Youngchol, Park, Yunjin, Moon, Junhyung, Jeong, Dongyoung, Park, Wanjin
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

Responsible AI Technical Report
por: KT, et al.
Publicado: (2025)

Adaptive Task Vectors for Large Language Models
por: Kang, Joonseong, et al.
Publicado: (2025)

SentGuard: Sentence-Level Streaming Guardrails for Large Language Models
por: Yu, Jiaqi, et al.
Publicado: (2026)

MrGuard: A Multilingual Reasoning Guardrail for Universal LLM Safety
por: Yang, Yahan, et al.
Publicado: (2025)

ConsisGuard: Aligning Safety Deliberation with Policy Enforcement in LLM Guardrails
por: Wang, Yan, et al.
Publicado: (2026)

BehaviorSFT: Behavioral Token Conditioning for Clinical Agents Across the Proactivity Spectrum
por: Kim, Yubin, et al.
Publicado: (2025)

EMBGuard: Constructing Hazard-Aware Guardrails for Safe Planning in Embodied Agents
por: Choi, Dongwook, et al.
Publicado: (2026)

DuoGuard: A Two-Player RL-Driven Framework for Multilingual LLM Guardrails
por: Deng, Yihe, et al.
Publicado: (2025)

Beyond Task-Oriented and Chitchat Dialogues: Proactive and Transition-Aware Conversational Agents
por: Yoon, Yejin, et al.
Publicado: (2025)

Task-Aware LoRA Adapter Composition via Similarity Retrieval in Vector Databases
por: Adsul, Riya, et al.
Publicado: (2026)

Beyond Two-Stage Training: Cooperative SFT and RL for LLM Reasoning
por: Chen, Liang, et al.
Publicado: (2025)

Beyond One Path: Evaluating and Enhancing Divergent Thinking in Interactive LLM Agents
por: Park, Jihyeong, et al.
Publicado: (2026)

KLAAD: Refining Attention Mechanisms to Reduce Societal Bias in Generative Language Models
por: Kim, Seorin, et al.
Publicado: (2025)

Rare-to-Frequent: Unlocking Compositional Generation Power of Diffusion Models on Rare Concepts with LLM Guidance
por: Park, Dongmin, et al.
Publicado: (2024)

InstaTrans: An Instruction-Aware Translation Framework for Non-English Instruction Datasets
por: Kim, Yungi, et al.
Publicado: (2024)

Does a Large Language Model Really Speak in Human-Like Language?
por: Park, Mose, et al.
Publicado: (2025)

Prefixing Attention Sinks can Mitigate Activation Outliers for Large Language Model Quantization
por: Son, Seungwoo, et al.
Publicado: (2024)

Guarding the Guardrails: A Taxonomy-Driven Approach to Jailbreak Detection
por: Giarrusso, Francesco, et al.
Publicado: (2025)

LitE-SQL: A Lightweight and Efficient Text-to-SQL Framework with Vector-based Schema Linking and Execution-Guided Self-Correction
por: Piao, Shengmin, et al.
Publicado: (2025)

Spread Preference Annotation: Direct Preference Judgment for Efficient LLM Alignment
por: Kim, Dongyoung, et al.
Publicado: (2024)

Benchmarking LLM Guardrails in Handling Multilingual Toxicity
por: Yang, Yahan, et al.
Publicado: (2024)

WebGuard: Building a Generalizable Guardrail for Web Agents
por: Zheng, Boyuan, et al.
Publicado: (2025)

LLM-Enhanced Linear Autoencoders for Recommendation
por: Moon, Jaewan, et al.
Publicado: (2025)

EnSToM: Enhancing Dialogue Systems with Entropy-Scaled Steering Vectors for Topic Maintenance
por: Suh, Heejae, et al.
Publicado: (2025)

Maximizing Local Entropy Where It Matters: Prefix-Aware Localized LLM Unlearning
por: Zhai, Naixin, et al.
Publicado: (2026)

Beyond Line-Level Filtering for the Pretraining Corpora of LLMs
por: Park, Chanwoo, et al.
Publicado: (2025)

Label Words as Local Task Vectors in In-Context Learning
por: Zheng, Bowen, et al.
Publicado: (2024)

ThinkGuard: Deliberative Slow Thinking Leads to Cautious Guardrails
por: Wen, Xiaofei, et al.
Publicado: (2025)

UniGuard: Towards Universal Safety Guardrails for Jailbreak Attacks on Multimodal Large Language Models
por: Oh, Sejoon, et al.
Publicado: (2024)

ConceptGuard: Neuro-Symbolic Safety Guardrails via Sparse Interpretable Jailbreak Concepts
por: Aswal, Darpan, et al.
Publicado: (2025)

OmniGuard: Unified Omni-Modal Guardrails with Deliberate Reasoning
por: Zhu, Boyu, et al.
Publicado: (2025)

Beyond the Final Answer: Evaluating the Reasoning Trajectories of Tool-Augmented Agents
por: Kim, Wonjoong, et al.
Publicado: (2025)

Moral Outrage Shapes Commitments Beyond Attention: Multimodal Moral Emotions on YouTube in Korea and the US
por: Park, Seongchan, et al.
Publicado: (2026)

Linguistics-Aware Non-Distortionary LLM Watermarking
por: Park, Shinwoo, et al.
Publicado: (2026)

Task Vectors, Learned Not Extracted: Performance Gains and Mechanistic Insight
por: Yang, Haolin, et al.
Publicado: (2025)

InjecGuard: Benchmarking and Mitigating Over-defense in Prompt Injection Guardrail Models
por: Li, Hao, et al.
Publicado: (2024)

SALSA: Speech Aware LLM Adaptation via Learned Steering Activation Vectors
por: Yegorova, Yekaterina, et al.
Publicado: (2026)

PrefixMemory-Tuning: Modernizing Prefix-Tuning by Decoupling the Prefix from Attention
por: Wang, Haonan, et al.
Publicado: (2025)

Designing Role Vectors to Improve LLM Inference Behaviour
por: Potertì, Daniele, et al.
Publicado: (2025)

Advancing Beyond Identification: Multi-bit Watermark for Large Language Models
por: Yoo, KiYoon, et al.
Publicado: (2023)