paper-with-me

Papers

Dynamic Multi-Reward Weighting for Multi-Style Controllable Generation

2024-02-21 · Karin de Langis, Ryan Koo, Dongyeop Kang

Textual style expresses a diverse set of information, including interpersonal dynamics (e.g., formality) and the author's emotions or attitudes (e.g., disgust). An open question is how language models can be explicitly controlled so that they weave together target styles when generating text: for example, to produce text that is both negative and non-toxic. One approach to such controlled generation is multi-objective reinforcement learning (RL), but how best to combine multiple objectives in a reward function is an open question. In this paper, we investigate various formulations of multi-style rewards, including calibrated outputs from discriminators and dynamic weighting by discriminator gradient magnitudes. We find that our proposed dynamic weighting outperforms static weighting approaches with respect to style control while maintaining linguistic quality, and we explore its effectiveness in 2- and 3-style control.

📄 PDF Abstract BibTeX arXiv:2402.14146

Code (1)

minnesotanlp/dynamic-multi-reward-weighting 공식 구현 pytorch

Tasks

Multi-Objective Reinforcement LearningReinforcement Learning (RL)

Methods 이 논문이 사용한 방법론

SET Dynamic Sparse Training method where weight mask is updated randomly periodically

Similar Papers 제목 키워드 기반

SAW: Stage-Aware Dynamic Weighting for Multi-Objective Reinforcement Learning in Large Language Models

2026-06-05 · Yuchen He, Baolong Bi, Shenghua Liu, Huaming Liao 외 arxiv

Although multi-objective reinforcement learning (MORL) is central to aligning large language models with complex human preferences, the prevailing practice of static weighted summation overlooks a more fundamental phenom…

Reinforcement LearningText Summarization

Learning to Optimize Multi-Objective Alignment Through Dynamic Reward Weighting

2025-09-14 · Yining Lu, Zilong Wang, Shiyang Li, Xin Liu 외 arxiv

Prior work in multi-objective reinforcement learning typically uses linear reward scalarization with fixed weights, which provably fails to capture non-convex Pareto fronts and thus yields suboptimal results. This limita…

Reinforcement Learning

Truncate Bad, Upweight Good: BoN-Style Distillation via Rank-Based Classification

2026-08-20 · Yarin Bar, Yaniv Romano arxiv

Inference-time selection methods, such as Best-of-N, improve generation by sampling a pool of candidates and selecting the top-ranked completion according to a reward model. Distillation seeks to amortize this procedure …

StyleVAR: Controllable Image Style Transfer via Visual Autoregressive Modeling

2026-04-22 · Liqi Jing, Dingming Zhang, Peinian Li, Lichen Zhu 외 arxiv

We build on the Visual Autoregressive Modeling (VAR) framework and formulate style transfer as conditional discrete sequence modeling in a learned latent space. Images are decomposed into multi-scale representations and …

Style Transfer

TIAR: Trajectory-Informed Advantage Reweighting for LLM Abstention Learning

2026-05-25 · Muyu Pan, Shu Zhao, Nan Zhang, Philip Shin 외 arxiv

This paper investigates large language model (LLM) abstention learning, specifically using ternary reward, which incentivize truthfulness in large language models. This paper extends that idea by moving from a ternary re…