Dense Reward for Free in Reinforcement Learning from Human Feedback

• URL: http://arxiv.org/abs/2402.00782v1
• Date: Thu, 1 Feb 2024 17:10:35 GMT
• Title: Dense Reward for Free in Reinforcement Learning from Human Feedback
• Authors: Alex J. Chan, Hao Sun, Samuel Holt, Mihaela van der Schaar
• Abstract summary: We leverage the fact that the reward model contains more information than just its scalar output. We use these attention weights to redistribute the reward along the whole completion. Empirically, we show that it stabilises training, accelerates the rate of learning, and, in practical cases, may lead to better local optima.
• Score: 64.92448888346125
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Reinforcement Learning from Human Feedback (RLHF) has been credited as the key advance that has allowed Large Language Models (LLMs) to effectively follow instructions and produce useful assistance. Classically, this involves generating completions from the LLM in response to a query before using a separate reward model to assign a score to the full completion. As an auto-regressive process, the LLM has to take many "actions" (selecting individual tokens) and only receives a single, sparse reward at the end of an episode, a setup that is known to be difficult to optimise in traditional reinforcement learning. In this work we leverage the fact that the reward model contains more information than just its scalar output, in particular, it calculates an attention map over tokens as part of the transformer architecture. We use these attention weights to redistribute the reward along the whole completion, effectively densifying the signal and highlighting the most important tokens, all without incurring extra computational cost or requiring any additional modelling. We demonstrate that, theoretically, this approach is equivalent to potential-based reward shaping, ensuring that the optimal policy remains unchanged. Empirically, we show that it stabilises training, accelerates the rate of learning, and, in practical cases, may lead to better local optima.

Related papers

• Approximated Variational Bayesian Inverse Reinforcement Learning for Large Language Model Alignment [0.618727087412292]
  The alignment of large language models (LLMs) is crucial for generating helpful and harmless content. Existing approaches leverage preference-based human feedback data to learn the reward function. We propose a novel training objective, Approximated Variational Alignment (AVA), to perform LLM alignment through Approximated Variational Reward Learning (AVRIL)
  arXiv  Detail & Related papers  (2024-11-14T10:37:34Z)
• R3HF: Reward Redistribution for Enhancing Reinforcement Learning from Human Feedback [25.27230140274847]
  Reinforcement learning from human feedback (RLHF) provides a paradigm for aligning large language models (LLMs) with human preferences. This paper proposes a novel reward redistribution method called R3HF, which facilitates a more fine-grained, token-level reward allocation.
  arXiv  Detail & Related papers  (2024-11-13T02:45:21Z)
• Assessing the Zero-Shot Capabilities of LLMs for Action Evaluation in RL [14.091146805312636]
  Credit assignment problem is a central challenge in Reinforcement Learning (RL) Credit Assignment with Language Models (CALM) is a novel approach to automate credit assignment via reward shaping and options discovery. Preliminary results indicate that the knowledge of Large Language Models is a promising prior for credit assignment in RL.
  arXiv  Detail & Related papers  (2024-09-19T14:08:09Z)
• Efficient Preference-based Reinforcement Learning via Aligned Experience Estimation [37.36913210031282]
  Preference-based reinforcement learning (PbRL) has shown impressive capabilities in training agents without reward engineering. We propose SEER, an efficient PbRL method that integrates label smoothing and policy regularization techniques.
  arXiv  Detail & Related papers  (2024-05-29T01:49:20Z)
• Countering Reward Over-optimization in LLM with Demonstration-Guided Reinforcement Learning [49.87923965553233]
  Reinforcement Learning can lead to reward over-optimization in large language models. We introduce the Reward from Demonstration (RCfD) to recalibrate the reward objective. We show that RCfD achieves comparable performance to carefully tuned baselines while mitigating ROO.
  arXiv  Detail & Related papers  (2024-04-30T09:57:21Z)
• Deep Reinforcement Learning from Hierarchical Preference Design [99.46415116087259]
  This paper shows by exploiting certain structures, one can ease the reward design process. We propose a hierarchical reward modeling framework -- HERON for scenarios: (I) The feedback signals naturally present hierarchy; (II) The reward is sparse, but with less important surrogate feedback to help policy learning.
  arXiv  Detail & Related papers  (2023-09-06T00:44:29Z)
• Basis for Intentions: Efficient Inverse Reinforcement Learning using Past Experience [89.30876995059168]
  inverse reinforcement learning (IRL) -- inferring the reward function of an agent from observing its behavior. This paper addresses the problem of IRL -- inferring the reward function of an agent from observing its behavior.
  arXiv  Detail & Related papers  (2022-08-09T17:29:49Z)
• SURF: Semi-supervised Reward Learning with Data Augmentation for Feedback-efficient Preference-based Reinforcement Learning [168.89470249446023]
  We present SURF, a semi-supervised reward learning framework that utilizes a large amount of unlabeled samples with data augmentation. In order to leverage unlabeled samples for reward learning, we infer pseudo-labels of the unlabeled samples based on the confidence of the preference predictor. Our experiments demonstrate that our approach significantly improves the feedback-efficiency of the preference-based method on a variety of locomotion and robotic manipulation tasks.
  arXiv  Detail & Related papers  (2022-03-18T16:50:38Z)
• On the Theory of Reinforcement Learning with Once-per-Episode Feedback [120.5537226120512]
  We introduce a theory of reinforcement learning in which the learner receives feedback only once at the end of an episode. This is arguably more representative of real-world applications than the traditional requirement that the learner receive feedback at every time step.
  arXiv  Detail & Related papers  (2021-05-29T19:48:51Z)
• Reward prediction for representation learning and reward shaping [0.8883733362171032]
  We propose learning a state representation in a self-supervised manner for reward prediction. We augment the training of out-of-the-box RL agents by shaping the reward using our reward predictor during policy learning.
  arXiv  Detail & Related papers  (2021-05-07T11:29:32Z)

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.