English

Leftover Lunch: Advantage-based Offline Reinforcement Learning for Language Models

Computation and Language 2024-04-23 v5

Abstract

Reinforcement Learning with Human Feedback (RLHF) is the most prominent method for Language Model (LM) alignment. However, RLHF is an unstable and data-hungry process that continually requires new high-quality LM-generated data for finetuning. We introduce Advantage-Leftover Lunch RL (A-LoL), a new class of offline policy gradient algorithms that enable RL training on any pre-existing data. By assuming the entire LM output sequence as a single action, A-LoL allows incorporating sequence-level classifiers or human-designed scoring functions as rewards. Subsequently, by using LM's value estimate, A-LoL only trains on positive advantage (leftover) data points, making it resilient to noise. Overall, A-LoL is an easy-to-implement, sample-efficient, and stable LM training recipe. We demonstrate the effectiveness of A-LoL and its variants with a set of four different language generation tasks. We compare against both online RL (PPO) and recent preference-based (DPO, PRO) and reward-based (GOLD) offline RL baselines. On the commonly-used RLHF benchmark, Helpful and Harmless Assistant (HHA), LMs trained with A-LoL methods achieve the highest diversity while also being rated more safe and helpful than the baselines according to humans. Additionally, in the remaining three tasks, A-LoL could optimize multiple distinct reward functions even when using noisy or suboptimal training data. We also release our experimental code. https://github.com/abaheti95/LoL-RL

Keywords

Cite

@article{arxiv.2305.14718,
  title  = {Leftover Lunch: Advantage-based Offline Reinforcement Learning for Language Models},
  author = {Ashutosh Baheti and Ximing Lu and Faeze Brahman and Ronan Le Bras and Maarten Sap and Mark Riedl},
  journal= {arXiv preprint arXiv:2305.14718},
  year   = {2024}
}

Comments

published at ICLR 2024

R2 v1 2026-06-28T10:43:58.575Z