Cover Image

Safety Recovery in Reasoning Models Is Only a Few Early Steering Steps Away

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Ghosal, Soumya Suvra, Chakraborty, Souradip, Singh, Vaibhav, Huang, Furong, Manocha, Dinesh, Bedi, Amrit Singh
Format:	Preprint
Published:	2026
Subjects:	Computation and Language Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Does Thinking More always Help? Mirage of Test-Time Scaling in Reasoning Models
by: Ghosal, Soumya Suvra, et al.
Published: (2025)

Bounded Rationality for LLMs: Satisficing Alignment at Inference-Time
by: Chehade, Mohamad, et al.
Published: (2025)

Transfer Q Star: Principled Decoding for LLM Alignment
by: Chakraborty, Souradip, et al.
Published: (2024)

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

Relic: Enhancing Reward Model Generalization for Low-Resource Indic Languages with Few-Shot Examples
by: Ghosal, Soumya Suvra, et al.
Published: (2025)

Code Comprehension then Auditing for Unsupervised LLM Evaluation
by: Patel, Bhrij, et al.
Published: (2024)

PromptRefine: Enhancing Few-Shot Performance on Low-Resource Indic Languages with Example Selection from Related Example Banks
by: Ghosal, Soumya Suvra, et al.
Published: (2024)

MaxMin-RLHF: Alignment with Diverse Human Preferences
by: Chakraborty, Souradip, et al.
Published: (2024)

Collab: Controlled Decoding using Mixture of Agents for LLM Alignment
by: Chakraborty, Souradip, et al.
Published: (2025)

Uncertainty-Aware Answer Selection for Improved Reasoning in Multi-LLM Systems
by: Agrawal, Aakriti, et al.
Published: (2025)

Engagement Undermines Safety: How Stereotypes and Toxicity Shape Humor in Language Models
by: Dogra, Atharvan, et al.
Published: (2025)

Multi-LLM QA with Embodied Exploration
by: Patel, Bhrij, et al.
Published: (2024)

LIAR: Leveraging Inference Time Alignment (Best-of-N) to Jailbreak LLMs in Seconds
by: Beetham, James, et al.
Published: (2024)

SAIL: Self-Improving Efficient Online Alignment of Large Language Models
by: Ding, Mucong, et al.
Published: (2024)

KITE: Kernelized and Information Theoretic Exemplars for In-Context Learning
by: Singh, Vaibhav, et al.
Published: (2025)

PARL: A Unified Framework for Policy Alignment in Reinforcement Learning from Human Feedback
by: Chakraborty, Souradip, et al.
Published: (2023)

IntCoOp: Interpretability-Aware Vision-Language Prompt Tuning
by: Ghosal, Soumya Suvra, et al.
Published: (2024)

REBEL: Reward Regularization-Based Approach for Robotic Reinforcement Learning from Human Feedback
by: Chakraborty, Souradip, et al.
Published: (2023)

Why Pass@k Optimization Can Degrade Pass@1: Prompt Interference in LLM Post-training
by: Barakat, Anas, et al.
Published: (2026)

Beyond Text: Utilizing Vocal Cues to Improve Decision Making in LLMs for Robot Navigation Tasks
by: Sun, Xingpeng, et al.
Published: (2024)

On the Global Optimality of Policy Gradient Methods in General Utility Reinforcement Learning
by: Barakat, Anas, et al.
Published: (2024)

What Drives Representation Steering? A Mechanistic Case Study on Steering Refusal
by: Cheng, Stephen, et al.
Published: (2026)

On the Vulnerability of LLM/VLM-Controlled Robotics
by: Wu, Xiyang, et al.
Published: (2024)

RL with Learnable Textual Feedback: A Bilevel Approach
by: Singh, Utsav, et al.
Published: (2026)

PROPS: Progressively Private Self-alignment of Large Language Models
by: Teku, Noel, et al.
Published: (2025)

DMCA: Dense Multi-agent Navigation using Attention and Communication
by: Arul, Senthil Hariharan, et al.
Published: (2022)

Draft-Conditioned Constrained Decoding for Structured Generation in LLMs
by: Reddy, Avinash, et al.
Published: (2026)

BalancedDPO: Adaptive Multi-Metric Alignment
by: Tamboli, Dipesh, et al.
Published: (2025)

VARP: Reinforcement Learning from Vision-Language Model Feedback with Agent Regularized Preferences
by: Singh, Anukriti, et al.
Published: (2025)

On the Role of Feedback in Test-Time Scaling of Agentic AI Workflows
by: Chakraborty, Souradip, et al.
Published: (2025)

Align-Pro: A Principled Approach to Prompt Optimization for LLM Alignment
by: Trivedi, Prashant, et al.
Published: (2025)

MIRA: Towards Mitigating Reward Hacking in Inference-Time Alignment of T2I Diffusion Models
by: Zhai, Kevin, et al.
Published: (2025)

Agentic Critical Training
by: Liu, Weize, et al.
Published: (2026)

Personalized Embodied Navigation for Portable Object Finding
by: Dorbala, Vishnu Sashank, et al.
Published: (2024)

AI Cap-and-Trade: Efficiency Incentives for Accessibility and Sustainability
by: Bornstein, Marco, et al.
Published: (2026)

LANTERN: LLM-Augmented Neurosymbolic Transfer with Experience-Gated Reasoning Networks
by: Alinejad, Mahyar, et al.
Published: (2026)

Enhancing Deep Neural Network Reliability with Refinement and Calibration
by: Hebbalaguppe, Ramya, et al.
Published: (2026)

Rethinking Adversarial Policies: A Generalized Attack Formulation and Provable Defense in RL
by: Liu, Xiangyu, et al.
Published: (2023)

Learning Multi-Robot Coordination through Locality-Based Factorized Multi-Agent Actor-Critic Algorithm
by: Shek, Chak Lam, et al.
Published: (2025)

Self-Evolved Preference Optimization for Enhancing Mathematical Reasoning in Small Language Models
by: Singh, Joykirat, et al.
Published: (2025)