Imagen de Portada

Reinforcement Learning with Semantic Rewards Enables Low-Resource Language Expansion without Alignment Tax

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Su, Zeli, Zhang, Ziyin, Liu, Zhou, Song, Xuexian, Xu, Zhankai, Zheng, Longfei, Zhang, Xiaolu, Fu, Rong, Xu, Guixian, Zhang, Wentao
Formato:	Preprint
Publicado:	2026
Materias:	Computation and Language Machine Learning
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Source-Grounded Semantic Reinforcement Learning for Low-Resource Target-Language Generation
por: Su, Zeli, et al.
Publicado: (2026)

FTibSuite: A Comprehensive Resource Suite for Tibetan Vision-Language Modeling
por: Xu, Guixian, et al.
Publicado: (2026)

The Curse of Helpfulness: Inverse Scaling Law in Robustness to Distractor Instructions via DistractionIF
por: Su, Zeli, et al.
Publicado: (2026)

SHRP: Specialized Head Routing and Pruning for Efficient Encoder Compression
por: Su, Zeli, et al.
Publicado: (2025)

Multilingual Encoder Knows more than You Realize: Shared Weights Pretraining for Extremely Low-Resource Languages
por: Su, Zeli, et al.
Publicado: (2025)

CMHG: A Dataset and Benchmark for Headline Generation of Minority Languages in China
por: Xu, Guixian, et al.
Publicado: (2025)

Group Symmetry Enables Faster Optimization in Inverse Problems
por: Tang, Junqi, et al.
Publicado: (2025)

Multilingual LLMs Inherently Reward In-Language Time-Sensitive Semantic Alignment for Low-Resource Languages
por: Bajpai, Ashutosh, et al.
Publicado: (2024)

Pareto-Guided Optimal Transport for Multi-Reward Alignment
por: Ba, Ying, et al.
Publicado: (2026)

Online Merging Optimizers for Boosting Rewards and Mitigating Tax in Alignment
por: Lu, Keming, et al.
Publicado: (2024)

Mitigating the Alignment Tax of RLHF
por: Lin, Yong, et al.
Publicado: (2023)

Safety Alignment as Continual Learning: Mitigating the Alignment Tax via Orthogonal Gradient Projection
por: Sun, Guanglong, et al.
Publicado: (2026)

FuRL: Visual-Language Models as Fuzzy Rewards for Reinforcement Learning
por: Fu, Yuwei, et al.
Publicado: (2024)

Policy Expansion for Bridging Offline-to-Online Reinforcement Learning
por: Zhang, Haichao, et al.
Publicado: (2023)

Anatomy-Aware Low-Dose CT Denoising via Pretrained Vision Models and Semantic-Guided Contrastive Learning
por: Wang, Runze, et al.
Publicado: (2025)

Focal Reward: Balanced Reinforcement Learning under Rubric-Based Rewards
por: Huang, Yu, et al.
Publicado: (2026)

The Hallucination Tax of Reinforcement Finetuning
por: Song, Linxin, et al.
Publicado: (2025)

Taxes without Taxpayers: The Invisibility of Taxes in Chile
por: Andrés Biehl
Publicado: (2019)

Zeroth-Order Policy Gradient for Reinforcement Learning from Human Feedback without Reward Inference
por: Zhang, Qining, et al.
Publicado: (2024)

Does Climate Finance Accelerate Progress Toward SDG 7? Evidence From Low‐and Middle‐Income Countries
por: Junyi Tian, et al.
Publicado: (2026)

Enhancing Reward Models for High-quality Image Generation: Beyond Text-Image Alignment
por: Ba, Ying, et al.
Publicado: (2025)

Prototypical Progressive Alignment and Reweighting for Generalizable Semantic Segmentation
por: Zhang, Yuhang, et al.
Publicado: (2025)

MERIT: Multilingual Expert-Reward Informed Tuning for Chinese-Centric Low-Resource Machine Translation
por: Lu, Zhixiang, et al.
Publicado: (2026)

Constraints as Rewards: Reinforcement Learning for Robots without Reward Functions
por: Ishihara, Yu, et al.
Publicado: (2025)

Inverse Reinforcement Learning without an Optimal Demonstrator: A Feasible Reward Set Approach
por: Kim, Kihyun, et al.
Publicado: (2026)

RMB: Comprehensively Benchmarking Reward Models in LLM Alignment
por: Zhou, Enyu, et al.
Publicado: (2024)

What Is the Alignment Tax?
por: Young, Robin
Publicado: (2026)

Locating a shortest vector in certain $2$-dimensional lattices
por: Zou, Guixian
Publicado: (2026)

Fast Gradient Methods for Data-Consistent Local Super-Resolution of Medical Images
por: Tang, Junqi, et al.
Publicado: (2022)

A Comparative Study of Variational Autoencoders, Normalizing Flows, and Score-based Diffusion Models for Electrical Impedance Tomography
por: Wang, Huihui, et al.
Publicado: (2023)

Fast Equivariant Imaging: Acceleration for Unsupervised Learning via Augmented Lagrangian and Auxiliary PnP Denoisers
por: Xu, Guixian, et al.
Publicado: (2025)

A New Convergence Analysis of Plug-and-Play Proximal Gradient Descent Under Prior Mismatch
por: Xu, Guixian, et al.
Publicado: (2026)

Equivariant Test-Time Training with Operator Sketching for Imaging Inverse Problems
por: Xu, Guixian, et al.
Publicado: (2024)

Enhancing Electrical Impedance Tomography reconstruction using Learned Half-Quadratic Splitting Networks with Anderson Acceleration
por: Xu, Guixian, et al.
Publicado: (2023)

Conscious Gaze: Adaptive Attention Mechanisms for Hallucination Mitigation in Vision-Language Models
por: Bu, Weijue, et al.
Publicado: (2025)

NALA: an Effective and Interpretable Entity Alignment Method
por: Xu, Chuanhao, et al.
Publicado: (2024)

Joint Semantic Transmission and Resource Allocation for Intelligent Computation Task Offloading in MEC Systems
por: Zheng, Yuanpeng, et al.
Publicado: (2025)

Pay Attention Later: Isolating the Semantic Alignment Tax via Iterative Semantic Map Refinement
por: YILDIRIM, Alper, et al.
Publicado: (2025)

Pay Attention Later: Isolating the Semantic Alignment Tax via Iterative Semantic Map Refinement
por: YILDIRIM, Alper, et al.
Publicado: (2025)

Three Mechanisms of Feature Learning in a Linear Network
por: Xu, Yizhou, et al.
Publicado: (2024)