Image de couverture de livre

Breaking the Safety-Capability Tradeoff: Reinforcement Learning with Verifiable Rewards Maintains Safety Guardrails in LLMs

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Cho, Dongkyu Derek, Song, Huan, Chowdhury, Arijit Ghosh, An, Haotian, Wang, Yawei, Thekkanal, Rohit, Sokhandan, Negin, Keshava, Sharlina, Marlowe, Hannah
Format:	Preprint
Publié:	2025
Sujets:	Machine Learning Artificial Intelligence
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Learning from Generalization Patterns: An Evaluation-Driven Approach to Enhanced Data Augmentation for Fine-Tuning Small Language Models
par: Song, Huan, et autres
Publié: (2025)

Improved Few-Shot Image Classification Through Multiple-Choice Questions
par: Khullar, Dipika, et autres
Publié: (2024)

Implicit Updates for Average-Reward Temporal Difference Learning
par: Kim, Hwanwoo, et autres
Publié: (2025)

Automated Virtual Product Placement and Assessment in Images using Diffusion Models
par: Alam, Mohammad Mahmudul, et autres
Publié: (2024)

CR4T: Rewrite-Based Guardrails for Adolescent LLM Safety
par: An, Heajun, et autres
Publié: (2026)

Breaking Free: How to Hack Safety Guardrails in Black-Box Diffusion Models!
par: Kotyan, Shashank, et autres
Publié: (2024)

Building Effective Safety Guardrails in AI Education Tools
par: Clark, Hannah-Beth, et autres
Publié: (2025)

SGuard-v1: Safety Guardrail for Large Language Models
par: Lee, JoonHo, et autres
Publié: (2025)

Safety Guardrails for LLM-Enabled Robots
par: Ravichandran, Zachary, et autres
Publié: (2025)

AGrail: A Lifelong Agent Guardrail with Effective and Adaptive Safety Detection
par: Luo, Weidi, et autres
Publié: (2025)

The Verifier Tax: Horizon Dependent Safety Success Tradeoffs in Tool Using LLM Agents
par: Sah, Tanmay, et autres
Publié: (2026)

Test-Time Training Undermines Safety Guardrails
par: Antonelli, Simone, et autres
Publié: (2026)

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

A Lightweight Explainable Guardrail for Prompt Safety
par: Islam, Md Asiful, et autres
Publié: (2026)

Backtracking for Safety
par: Sel, Bilgehan, et autres
Publié: (2025)

The Safety-Privacy Tradeoff in Linear Bandits
par: Zibaie, Arghavan, et autres
Publié: (2025)

Fast Computer Model Calibration using Annealed and Transformed Variational Inference
par: Cho, Dongkyu Derek, et autres
Publié: (2022)

Bag of Tricks for Subverting Reasoning-based Safety Guardrails
par: Chen, Shuo, et autres
Publié: (2025)

Why Do Safety Guardrails Degrade Across Languages?
par: Zhang, Max, et autres
Publié: (2026)

PoseGuard: Pose-Guided Generation with Safety Guardrails
par: Wang, Kongxin, et autres
Publié: (2025)

When in Doubt, Cascade: Towards Building Efficient and Capable Guardrails
par: Nagireddy, Manish, et autres
Publié: (2024)

Generative Data Augmentation using LLMs improves Distributional Robustness in Question Answering
par: Chowdhury, Arijit Ghosh, et autres
Publié: (2023)

Lightweight Safety Guardrails Using Fine-tuned BERT Embeddings
par: Zheng, Aaron, et autres
Publié: (2024)

Safety Through Reasoning: An Empirical Study of Reasoning Guardrail Models
par: Sreedhar, Makesh Narsimhan, et autres
Publié: (2025)

Guardrails in Logit Space: Safety Token Regularization for LLM Alignment
par: Bach, Thong, et autres
Publié: (2026)

A Guardrail for Safety Preservation: When Safety-Sensitive Subspace Meets Harmful-Resistant Null-Space
par: Zhang, Bingjie, et autres
Publié: (2025)

RLNVR: Reinforcement Learning from Non-Verified Real-World Rewards
par: Krishnan, Rohit, et autres
Publié: (2025)

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

Belonging and Transnational Refugee Settlement
par: Marlowe, Jay
Publié: (2021)

Belonging and Transnational Refugee Settlement
par: Marlowe, Jay
Publié: (2021)

Cutting Loose.
par: Marlowe, John
Publié: (1994)

Dealing with the Evil Twins: Improving Random Augmentation by Addressing Catastrophic Forgetting of Diverse Augmentations
par: Cho, Dongkyu, et autres
Publié: (2025)

The Supportiveness-Safety Tradeoff in LLM Well-Being Agents
par: Lalwani, Himanshi, et autres
Publié: (2026)

Elucidating Optimal Reward-Diversity Tradeoffs in Text-to-Image Diffusion Models
par: Jena, Rohit, et autres
Publié: (2024)

PSG-Agent: Personality-Aware Safety Guardrail for LLM-based Agents
par: Wu, Yaozu, et autres
Publié: (2025)

SafeHarbor: Hierarchical Memory-Augmented Guardrail for LLM Agent Safety
par: Liu, Zhe, et autres
Publié: (2026)

Bridging the Safety Gap: A Guardrail Pipeline for Trustworthy LLM Inferences
par: Han, Shanshan, et autres
Publié: (2025)

ConsisGuard: Aligning Safety Deliberation with Policy Enforcement in LLM Guardrails
par: Wang, Yan, et autres
Publié: (2026)

MrGuard: A Multilingual Reasoning Guardrail for Universal LLM Safety
par: Yang, Yahan, et autres
Publié: (2025)

Auto-Tuning Safety Guardrails for Black-Box Large Language Models
par: Abdulkadir, Perry
Publié: (2025)