paper-with-me

Papers

Contrastive Policy Gradient: Aligning LLMs on sequence-level scores in a supervised-friendly fashion

2024-06-27 · Yannis Flet-Berliac, Nathan Grinsztajn, Florian Strub, Bill Wu, Eugene Choi, Chris Cremer, Arash Ahmadian, Yash Chandak, Mohammad Gheshlaghi Azar, Olivier Pietquin, Matthieu Geist

Reinforcement Learning (RL) has been used to finetune Large Language Models (LLMs) using a reward model trained from preference data, to better align with human judgment. The recently introduced direct alignment methods, which are often simpler, more stable, and computationally lighter, can more directly achieve this. However, these approaches cannot optimize arbitrary rewards, and the preference-based ones are not the only rewards of interest for LLMs (eg., unit tests for code generation or textual entailment for summarization, among others). RL-finetuning is usually done with a variation of policy gradient, which calls for on-policy or near-on-policy samples, requiring costly generations. We introduce Contrastive Policy Gradient, or CoPG, a simple and mathematically principled new RL algorithm that can estimate the optimal policy even from off-policy data. It can be seen as an off-policy policy gradient approach that does not rely on important sampling techniques and highlights the importance of using (the right) state baseline. We show this approach to generalize the direct alignment method IPO (identity preference optimization) and classic policy gradient. We experiment with the proposed CoPG on a toy bandit problem to illustrate its properties, as well as for finetuning LLMs on a summarization task, using a learned reward function considered as ground truth for the purpose of the experiments.

📄 PDF Abstract BibTeX arXiv:2406.19185

Code (0)

등록된 구현이 없습니다.

Tasks

Code GenerationNatural Language InferenceReinforcement Learning (RL)

Methods 이 논문이 사용한 방법론

ALIGN In the ALIGN method, visual and language representations are jointly trained from noisy image alt-text data. The image and text encoders are learned via contrastive loss…

Similar Papers 제목 키워드 기반

SPG: Sandwiched Policy Gradient for Masked Diffusion Language Models

2025-10-10 · Chenyu Wang, Paria Rashidinejad, DiJia Su, Song Jiang 외 arxiv

Diffusion large language models (dLLMs) are emerging as an efficient alternative to autoregressive models due to their ability to decode multiple tokens in parallel. However, aligning dLLMs with human preferences or task…

Reinforcement Learning

Gradient-Adaptive Policy Optimization: Towards Multi-Objective Alignment of Large Language Models

2025-07-02 · Chengao Li, Hanyu Zhang, Yunkun Xu, Hongyan Xue 외 arxiv

Reinforcement Learning from Human Feedback (RLHF) has emerged as a powerful technique for aligning large language models (LLMs) with human preferences. However, effectively aligning LLMs with diverse human preferences re…

Reinforcement Learning

Cal-DPO: Calibrated Direct Preference Optimization for Language Model Alignment

2024-12-19 · Teng Xiao, Yige Yuan, Huaisheng Zhu, Mingxiao Li 외

We study the problem of aligning large language models (LLMs) with human preference data. Contrastive preference optimization has shown promising results in aligning LLMs with available preference data by optimizing the …

Language ModelingLanguage Modelling

Nash Learning from Human Feedback

2023-12-01 · Rémi Munos, Michal Valko, Daniele Calandriello, Mohammad Gheshlaghi Azar 외

Reinforcement learning from human feedback (RLHF) has emerged as the main paradigm for aligning large language models (LLMs) with human preferences. Typically, RLHF involves the initial step of learning a reward model fr…

Text Summarization

Simple Policy Gradients for Reasoning with Diffusion Language Models

2025-10-05 · Anthony Zhan arxiv

Diffusion large language models (dLLMs) represent a promising alternative to autoregressive LLMs; however, the lack of effective post-training techniques, including reinforcement learning (RL), remains a key challenge fo…

Reinforcement Learning