Copertina

LLM Safety Alignment is Divergence Estimation in Disguise

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Haldar, Rajdeep, Wang, Ziyi, Song, Qifan, Lin, Guang, Xing, Yue
Natura:	Preprint
Pubblicazione:	2025
Soggetti:	Machine Learning Artificial Intelligence Computers and Society
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

f-GRPO and Beyond: Divergence-Based Reinforcement Learning Algorithms for General LLM Alignment
di: Haldar, Rajdeep, et al.
Pubblicazione: (2026)

Adversarial Vulnerability as a Consequence of On-Manifold Inseparibility
di: Haldar, Rajdeep, et al.
Pubblicazione: (2024)

Effect of Ambient-Intrinsic Dimension Gap on Adversarial Vulnerability
di: Haldar, Rajdeep, et al.
Pubblicazione: (2024)

SQS: Bayesian DNN Compression through Sparse Quantized Sub-distributions
di: Wang, Ziyi, et al.
Pubblicazione: (2025)

Moral Alignment for LLM Agents
di: Tennant, Elizaveta, et al.
Pubblicazione: (2024)

Distributional Open-Ended Evaluation of LLM Cultural Value Alignment Based on Value Codebook
di: Lee, Jaehyeok, et al.
Pubblicazione: (2026)

Foundational Challenges in Assuring Alignment and Safety of Large Language Models
di: Anwar, Usman, et al.
Pubblicazione: (2024)

Superficial Safety Alignment Hypothesis
di: Li, Jianwei, et al.
Pubblicazione: (2024)

When Style Breaks Safety: Defending LLMs Against Superficial Style Alignment
di: Xiao, Yuxin, et al.
Pubblicazione: (2025)

FFB: A Fair Fairness Benchmark for In-Processing Group Fairness Methods
di: Han, Xiaotian, et al.
Pubblicazione: (2023)

Attacks, Defenses and Evaluations for LLM Conversation Safety: A Survey
di: Dong, Zhichen, et al.
Pubblicazione: (2024)

Evaluating LLM Agent Adherence to Hierarchical Safety Principles: A Lightweight Benchmark for Probing Foundational Controllability Components
di: Potham, Ram
Pubblicazione: (2025)

When the Domain Expert Has No Time and the LLM Developer Has No Clinical Expertise: Real-World Lessons from LLM Co-Design in a Safety-Net Hospital
di: Kothari, Avni, et al.
Pubblicazione: (2025)

Alignment as Jurisprudence
di: Caputo, Nicholas
Pubblicazione: (2026)

Generalized Discrete Diffusion with Self-Correction
di: Wang, Linxuan, et al.
Pubblicazione: (2026)

International AI Safety Report
di: Bengio, Yoshua, et al.
Pubblicazione: (2025)

AI Alignment at Your Discretion
di: Buyl, Maarten, et al.
Pubblicazione: (2025)

Pluralistic Alignment Over Time
di: Klassen, Toryn Q., et al.
Pubblicazione: (2024)

Why Conclusions Diverge from the Same Observations: Formalizing World-Model Non-Identifiability via an Inference
di: Takahashi, Toru
Pubblicazione: (2026)

An Approach to Technical AGI Safety and Security
di: Shah, Rohin, et al.
Pubblicazione: (2025)

Introduction to AI Safety, Ethics, and Society
di: Hendrycks, Dan
Pubblicazione: (2024)

Safety challenges of AI in medicine in the era of large language models
di: Wang, Xiaoye, et al.
Pubblicazione: (2024)

Open Problems in Machine Unlearning for AI Safety
di: Barez, Fazl, et al.
Pubblicazione: (2025)

Unified Uncertainty Estimation for Cognitive Diagnosis Models
di: Wang, Fei, et al.
Pubblicazione: (2024)

MANTA: Multi-turn Assessment for Nonhuman Thinking & Alignment
di: Lu, Allen, et al.
Pubblicazione: (2026)

Defining and Evaluating Physical Safety for Large Language Models
di: Tang, Yung-Chen, et al.
Pubblicazione: (2024)

An AI System Evaluation Framework for Advancing AI Safety: Terminology, Taxonomy, Lifecycle Mapping
di: Xia, Boming, et al.
Pubblicazione: (2024)

Being Considerate as a Pathway Towards Pluralistic Alignment for Agentic AI
di: Alamdari, Parand A., et al.
Pubblicazione: (2024)

On Catastrophic Inheritance of Large Foundation Models
di: Chen, Hao, et al.
Pubblicazione: (2024)

AI Companies Should Report Pre- and Post-Mitigation Safety Evaluations
di: Bowen, Dillon, et al.
Pubblicazione: (2025)

Safetywashing: Do AI Safety Benchmarks Actually Measure Safety Progress?
di: Ren, Richard, et al.
Pubblicazione: (2024)

NaiAD: Initiate Data-Driven Research for LLM Advertising
di: Zhang, Yihang, et al.
Pubblicazione: (2026)

The Secret Agenda: LLMs Strategically Lie and Our Current Safety Tools Are Blind
di: DeLeeuw, Caleb, et al.
Pubblicazione: (2025)

Enhancing Safety in Reinforcement Learning with Human Feedback via Rectified Policy Optimization
di: Peng, Xiyue, et al.
Pubblicazione: (2024)

Decolonial AI Alignment: Openness, Viśe\d{s}a-Dharma, and Including Excluded Knowledges
di: Varshney, Kush R.
Pubblicazione: (2023)

The Alignment Trap: Complexity Barriers
di: Yao, Jasper
Pubblicazione: (2025)

Assessing Social Alignment: Do Personality-Prompted Large Language Models Behave Like Humans?
di: Zakazov, Ivan, et al.
Pubblicazione: (2024)

Predicting Human Mobility during Extreme Events via LLM-Enhanced Cross-City Learning
di: Tang, Yinzhou, et al.
Pubblicazione: (2025)

Align Once, Benefit Multilingually: Enforcing Multilingual Consistency for LLM Safety Alignment
di: Bu, Yuyan, et al.
Pubblicazione: (2026)

Chain-of-Thought Unfaithfulness as Disguised Accuracy
di: Bentham, Oliver, et al.
Pubblicazione: (2024)