Cover Image

Gradient-Controlled Decoding: A Safety Guardrail for LLMs with Dual-Anchor Steering

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Chiniya, Purva, Scaria, Kevin, Chaturvedi, Sagar
Format:	Preprint
Published:	2026
Subjects:	Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Embarrassingly Simple Unsupervised Aspect Based Sentiment Tuple Extraction
by: Scaria, Kevin, et al.
Published: (2024)

Bypassing Safety Guardrails in LLMs Using Humor
by: Cisneros-Velarde, Pedro
Published: (2025)

LipGER: Visually-Conditioned Generative Error Correction for Robust Automatic Speech Recognition
by: Ghosh, Sreyan, et al.
Published: (2024)

Guardrail Baselines for Unlearning in LLMs
by: Thaker, Pratiksha, et al.
Published: (2024)

Interpretable LLM Guardrails via Sparse Representation Steering
by: He, Zeqing, et al.
Published: (2025)

SteeringSafety: A Systematic Safety Evaluation Framework of Representation Steering in LLMs
by: Siu, Vincent, et al.
Published: (2025)

A Lightweight Explainable Guardrail for Prompt Safety
by: Islam, Md Asiful, et al.
Published: (2026)

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

Steering Multimodal Large Language Models Decoding for Context-Aware Safety
by: Liu, Zheyuan, et al.
Published: (2025)

Lightweight Safety Guardrails Using Fine-tuned BERT Embeddings
by: Zheng, Aaron, et al.
Published: (2024)

Do LLM Agents Mirror Socio-Cognitive Effects in Power-Asymmetric Conversations?
by: Vijjini, Anvesh Rao, et al.
Published: (2026)

Bag of Tricks for Subverting Reasoning-based Safety Guardrails
by: Chen, Shuo, et al.
Published: (2025)

Capturing Bias Diversity in LLMs
by: Gosavi, Purva Prasad, et al.
Published: (2024)

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

LLM-Based Persuasion Enables Guardrail Override in Frontier LLMs
by: Nogueira, Rodrigo, et al.
Published: (2026)

CR4T: Rewrite-Based Guardrails for Adolescent LLM Safety
by: An, Heajun, et al.
Published: (2026)

KV Cache Steering for Controlling Frozen LLMs
by: Belitsky, Max, et al.
Published: (2025)

Safety Through Reasoning: An Empirical Study of Reasoning Guardrail Models
by: Sreedhar, Makesh Narsimhan, et al.
Published: (2025)

AV-RIR: Audio-Visual Room Impulse Response Estimation
by: Ratnarajah, Anton, et al.
Published: (2023)

Why Do Safety Guardrails Degrade Across Languages?
by: Zhang, Max, et al.
Published: (2026)

Steer Model beyond Assistant: Controlling System Prompt Strength via Contrastive Decoding
by: Dong, Yijiang River, et al.
Published: (2026)

TWGuard: A Case Study of LLM Safety Guardrails for Localized Linguistic Contexts
by: Chu, Hua-Rong, et al.
Published: (2026)

Generalization or Memorization: Dynamic Decoding for Mode Steering
by: Zhang, Xuanming
Published: (2025)

Deep Research with Open-Domain Evaluation and Multi-Stage Guardrails for Safety
by: Huang, Wei-Chieh, et al.
Published: (2025)

Is Safer Better? The Impact of Guardrails on the Argumentative Strength of LLMs in Hate Speech Countering
by: Bonaldi, Helena, et al.
Published: (2024)

GeoSteer: Faithful Chain-of-Thought Steering via Latent Manifold Gradients
by: Kazama, Kentaro, et al.
Published: (2026)

Cutting Through the Noise: Boosting LLM Performance on Math Word Problems
by: Anantheswaran, Ujjwala, et al.
Published: (2024)

Aegis2.0: A Diverse AI Safety Dataset and Risks Taxonomy for Alignment of LLM Guardrails
by: Ghosh, Shaona, et al.
Published: (2025)

SteerConf: Steering LLMs for Confidence Elicitation
by: Zhou, Ziang, et al.
Published: (2025)

SGuard-v1: Safety Guardrail for Large Language Models
by: Lee, JoonHo, et al.
Published: (2025)

SafeRoute: Adaptive Model Selection for Efficient and Accurate Safety Guardrails in Large Language Models
by: Lee, Seanie, et al.
Published: (2025)

GrAInS: Gradient-based Attribution for Inference-Time Steering of LLMs and VLMs
by: Nguyen, Duy, et al.
Published: (2025)

FairSteer: Inference Time Debiasing for LLMs with Dynamic Activation Steering
by: Li, Yichen, et al.
Published: (2025)

CREST: Universal Safety Guardrails Through Cluster-Guided Cross-Lingual Transfer
by: Bansal, Lavish, et al.
Published: (2025)

Guiding Giants: Lightweight Controllers for Weighted Activation Steering in LLMs
by: Hegazy, Amr, et al.
Published: (2025)

Steering LLMs for Culturally Localized Generation
by: Khanuja, Simran, et al.
Published: (2026)

CoSteer: Collaborative Decoding-Time Personalization via Local Delta Steering
by: Lv, Hang, et al.
Published: (2025)

Forewarned is Forearmed: Pre-Synthesizing Jailbreak-like Instructions to Enhance LLM Safety Guardrail to Potential Attacks
by: Liu, Sheng, et al.
Published: (2025)

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

Steered Generation via Gradient Descent on Sparse Features
by: Bhattacharyya, Sumanta, et al.
Published: (2025)