Imagen de Portada

Provable Last-Iterate Convergence for Multi-Objective Safe LLM Alignment via Optimistic Primal-Dual

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Li, Yining, Ju, Peizhong, Shroff, Ness
Formato:	Preprint
Publicado:	2026
Materias:	Machine Learning Artificial Intelligence
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

How to Find the Exact Pareto Front for Multi-Objective MDPs?
por: Li, Yining, et al.
Publicado: (2024)

PSMGD: Periodic Stochastic Multi-Gradient Descent for Fast Multi-Objective Optimization
por: Xu, Mingjing, et al.
Publicado: (2024)

Broadening Target Distributions for Accelerated Diffusion Models via a Novel Analysis Approach
por: Liang, Yuchen, et al.
Publicado: (2024)

Learnable Chernoff Baselines for Inference-Time Alignment
por: Madhow, Sunil, et al.
Publicado: (2026)

Provably Convergent Primal-Dual DPO for Constrained LLM Alignment
por: Du, Yihan, et al.
Publicado: (2025)

Provably Efficient Multi-Objective Bandit Algorithms under Preference-Centric Customization
por: Cao, Linfeng, et al.
Publicado: (2025)

Theory on Score-Mismatched Diffusion Models and Zero-Shot Conditional Samplers
por: Liang, Yuchen, et al.
Publicado: (2024)

An LP-based Sampling Policy for Multi-Armed Bandits with Side-Observations and Stochastic Availability
por: Soni, Ashutosh, et al.
Publicado: (2026)

Near-Optimal Partially Observable Reinforcement Learning with Partial Online State Information
por: Shi, Ming, et al.
Publicado: (2023)

Optimistic Reinforcement Learning with Quantile Objectives
por: Alipour-Vaezi, Mohammad, et al.
Publicado: (2025)

BeST -- A Novel Source Selection Metric for Transfer Learning
por: Soni, Ashutosh, et al.
Publicado: (2025)

Last-Iterate Convergence of General Parameterized Policies in Constrained MDPs
por: Mondal, Washim Uddin, et al.
Publicado: (2024)

Multi-Step Alignment as Markov Games: An Optimistic Online Gradient Descent Approach with Convergence Guarantees
por: Wu, Yongtao, et al.
Publicado: (2025)

Adaptive Primal-Dual Method for Safe Reinforcement Learning
por: Chen, Weiqin, et al.
Publicado: (2024)

Can We Theoretically Quantify the Impacts of Local Updates on the Generalization Performance of Federated Learning?
por: Ju, Peizhong, et al.
Publicado: (2024)

Improving LLM General Preference Alignment via Optimistic Online Mirror Descent
por: Zhang, Yuheng, et al.
Publicado: (2025)

Theory on Mixture-of-Experts in Continual Learning
por: Li, Hongbo, et al.
Publicado: (2024)

Global Convergence for Average Reward Constrained MDPs with Primal-Dual Actor Critic Algorithm
por: Xu, Yang, et al.
Publicado: (2025)

Discrete Flow Matching for Offline-to-Online Reinforcement Learning
por: Khan, Fairoz Nower, et al.
Publicado: (2026)

Implicit Safe Set Algorithm for Provably Safe Reinforcement Learning
por: Zhao, Weiye, et al.
Publicado: (2024)

Absorb and Converge: Provable Convergence Guarantee for Absorbing Discrete Diffusion Models
por: Liang, Yuchen, et al.
Publicado: (2025)

Unlocking the Power of Rehearsal in Continual Learning: A Theoretical Perspective
por: Deng, Junze, et al.
Publicado: (2025)

SafeAdapt: Provably Safe Policy Updates in Deep Reinforcement Learning
por: Anisimov, Maksim, et al.
Publicado: (2026)

Discrete MeanFlow: One-Step Generation via Conditional Transition Kernels
por: Khan, Fairoz Nower, et al.
Publicado: (2026)

Closing the Gap: Achieving Global Convergence (Last Iterate) of Actor-Critic under Markovian Sampling with Neural Network Parametrization
por: Gaur, Mudit, et al.
Publicado: (2024)

Last-Iterate Convergent Policy Gradient Primal-Dual Methods for Constrained MDPs
por: Ding, Dongsheng, et al.
Publicado: (2023)

Safe and Balanced: A Framework for Constrained Multi-Objective Reinforcement Learning
por: Gu, Shangding, et al.
Publicado: (2024)

Optimistic Policy Regularization
por: Pham, Mai, et al.
Publicado: (2026)

Dynamic Model Predictive Shielding for Provably Safe Reinforcement Learning
por: Banerjee, Arko, et al.
Publicado: (2024)

Uniformly Safe RL with Objective Suppression for Multi-Constraint Safety-Critical Applications
por: Zhou, Zihan, et al.
Publicado: (2024)

Adaptive Alignment: Dynamic Preference Adjustments via Multi-Objective Reinforcement Learning for Pluralistic AI
por: Harland, Hadassah, et al.
Publicado: (2024)

PARM: Multi-Objective Test-Time Alignment via Preference-Aware Autoregressive Reward Model
por: Lin, Baijiong, et al.
Publicado: (2025)

Primal-Dual Spectral Representation for Off-policy Evaluation
por: Hu, Yang, et al.
Publicado: (2024)

Leveraging Analytic Gradients in Provably Safe Reinforcement Learning
por: Walter, Tim, et al.
Publicado: (2025)

Optimistic Rates for Learning from Label Proportions
por: Li, Gene, et al.
Publicado: (2024)

Enumerating Safe Regions in Deep Neural Networks with Provable Probabilistic Guarantees
por: Marzari, Luca, et al.
Publicado: (2023)

What is the Alignment Objective of GRPO?
por: Vojnovic, Milan, et al.
Publicado: (2025)

Property-driven Protein Inverse Folding With Multi-Objective Preference Alignment
por: Hou, Xiaoyang, et al.
Publicado: (2026)

Pareto Multi-Objective Alignment for Language Models
por: He, Qiang, et al.
Publicado: (2025)

STIMULUS: Achieving Fast Convergence and Low Sample Complexity in Stochastic Multi-Objective Learning
por: Liu, Zhuqing, et al.
Publicado: (2025)