The N+ Implementation Details of RLHF with PPO: A Case Study on TL;DR Summarization

• URL: http://arxiv.org/abs/2403.17031v1
• Date: Sun, 24 Mar 2024 02:59:27 GMT
• Title: The N+ Implementation Details of RLHF with PPO: A Case Study on TL;DR Summarization
• Authors: Shengyi Huang, Michael Noukhovitch, Arian Hosseini, Kashif Rasul, Weixun Wang, Lewis Tunstall,
• Abstract summary: This work is the first to openly reproduce the Reinforcement Learning from Human Feedback scaling behaviors reported in OpenAI's seminal TL;DR summarization work. We create an RLHF pipeline from scratch, enumerate over 20 key implementation details, and share key insights during the reproduction.
• Score: 8.911768677958753
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: This work is the first to openly reproduce the Reinforcement Learning from Human Feedback (RLHF) scaling behaviors reported in OpenAI's seminal TL;DR summarization work. We create an RLHF pipeline from scratch, enumerate over 20 key implementation details, and share key insights during the reproduction. Our RLHF-trained Pythia models demonstrate significant gains in response quality that scale with model size, with our 2.8B, 6.9B models outperforming OpenAI's released 1.3B checkpoint. We publicly release the trained model checkpoints and code to facilitate further research and accelerate progress in the field (\url{https://github.com/vwxyzjn/summarize_from_feedback_details}).

Related papers

• Provably Efficient RLHF Pipeline: A Unified View from Contextual Bandits [59.30310692855397]
  We propose a unified framework for the RLHF pipeline from the view of contextual bandits. We decompose the RLHF process into two distinct stages: (post-)training and deployment. We then develop novel algorithms for each stage, demonstrating significant improvements in both statistical and computational efficiency.
  arXiv  Detail & Related papers  (2025-02-11T02:36:01Z)
• Does RLHF Scale? Exploring the Impacts From Data, Model, and Method [83.53178716807776]
  This study explores the scaling properties of Reinforcement Learning from Human Feedback in Large Language Models. We analyze key components in the RLHF framework--model size, data composition, and inference budget--and their impacts on performance.
  arXiv  Detail & Related papers  (2024-12-08T17:19:48Z)
• How to Evaluate Reward Models for RLHF [51.31240621943791]
  We introduce a new benchmark for reward models that quantifies their ability to produce strong language models through RLHF (Reinforcement Learning from Human Feedback) We build a predictive model of downstream LLM performance by evaluating the reward model on proxy tasks. We launch an end-to-end RLHF experiment on a large-scale crowdsourced human preference platform to view real reward model downstream performance as ground truth.
  arXiv  Detail & Related papers  (2024-10-18T21:38:21Z)
• DogeRM: Equipping Reward Models with Domain Knowledge through Model Merging [65.41765072566287]
  We propose textbfDomain knowledtextbfge merged textbfReward textbfModel (DogeRM), a novel framework that integrates domain-specific knowledge into a general reward model by model merging.
  arXiv  Detail & Related papers  (2024-07-01T17:01:54Z)
• RLHF Workflow: From Reward Modeling to Online RLHF [79.83927049253924]
  We present the workflow of Online Iterative Reinforcement Learning from Human Feedback (RLHF) in this technical report. RLHF is widely reported to outperform its offline counterpart by a large margin in the recent large language model (LLM) literature. We show that supervised fine-tuning (SFT) and iterative RLHF can obtain state-of-the-art performance with fully open-source datasets.
  arXiv  Detail & Related papers  (2024-05-13T15:50:39Z)
• A Theoretical Framework for Partially Observed Reward-States in RLHF [39.41038579993645]
  We model reinforcement learning with partially observed reward-states (PORRL) We accommodate two kinds of feedback $-$ cardinal and dueling feedback. In both feedback settings, we show that our models and guarantees generalize and extend existing ones.
  arXiv  Detail & Related papers  (2024-02-05T18:38:55Z)
• A Long Way to Go: Investigating Length Correlations in RLHF [59.49656695716066]
  This paper demonstrates, on three diverse settings, that optimizing for response length is a significant factor behind RLHF. We find improvements in reward to largely be driven by increasing response length, instead of other features. Even a purely length-based reward reproduces most downstream RLHF improvements over supervised fine-tuned models.
  arXiv  Detail & Related papers  (2023-10-05T17:38:28Z)
• RRHF: Rank Responses to Align Language Models with Human Feedback without tears [69.68672043223249]
  InstructGPT implements RLHF through several stages, including Supervised Fine-Tuning (SFT), reward model training, and Proximal Policy Optimization (PPO) We propose a novel learning paradigm called RRHF, which scores sampled responses from different sources via a logarithm of conditional probabilities. We evaluate RRHF on the Helpful and Harmless dataset, demonstrating comparable alignment performance with PPO by reward model score and human labeling.
  arXiv  Detail & Related papers  (2023-04-11T15:53:40Z)
• Inverse Reinforcement Learning for Text Summarization [52.765898203824975]
  We introduce inverse reinforcement learning (IRL) as an effective paradigm for training abstractive summarization models. Experimental results across datasets in different domains demonstrate the superiority of our proposed IRL model for summarization over MLE and RL baselines.
  arXiv  Detail & Related papers  (2022-12-19T23:45:05Z)
• RewardsOfSum: Exploring Reinforcement Learning Rewards for Summarisation [7.0471949371778795]
  We propose two reward functions for the task of abstractive summarisation. The first function, referred to as RwB-Hinge, dynamically selects the samples for the gradient update. The second function, nicknamed RISK, leverages a small pool of strong candidates to inform the reward.
  arXiv  Detail & Related papers  (2021-06-08T03:30:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.