On Reinforcement Learning and Distribution Matching for Fine-Tuning Language Models with no Catastrophic Forgetting

• URL: http://arxiv.org/abs/2206.00761v1
• Date: Wed, 1 Jun 2022 20:54:41 GMT
• Title: On Reinforcement Learning and Distribution Matching for Fine-Tuning Language Models with no Catastrophic Forgetting
• Authors: Tomasz Korbak and Hady Elsahar and Germ\'an Kruszewski and Marc Dymetman
• Abstract summary: Two main paradigms have emerged to tackle this challenge: Reward Maximization (RM) and, more recently, Distribution Matching (DM) We show that methods such as KL-control developed for RM can also be construed as belonging to DM. We leverage connections between the two paradigms to import the concept of baseline into DM methods.
• Score: 5.5302127686575435
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The availability of large pre-trained models is changing the landscape of Machine Learning research and practice, moving from a training-from-scratch to a fine-tuning paradigm. While in some applications the goal is to "nudge" the pre-trained distribution towards preferred outputs, in others it is to steer it towards a different distribution over the sample space. Two main paradigms have emerged to tackle this challenge: Reward Maximization (RM) and, more recently, Distribution Matching (DM). RM applies standard Reinforcement Learning (RL) techniques, such as Policy Gradients, to gradually increase the reward signal. DM prescribes to first make explicit the target distribution that the model is fine-tuned to approximate. Here we explore the theoretical connections between the two paradigms, and show that methods such as KL-control developed for RM can also be construed as belonging to DM. We further observe that while DM differs from RM, it can suffer from similar training difficulties, such as high gradient variance. We leverage connections between the two paradigms to import the concept of baseline into DM methods. We empirically validate the benefits of adding a baseline on an array of controllable language generation tasks such as constraining topic, sentiment, and gender distributions in texts sampled from a language model. We observe superior performance in terms of constraint satisfaction, stability and sample efficiency.

Related papers

• Diffusing States and Matching Scores: A New Framework for Imitation Learning [16.941612670582522]
  Adversarial Imitation Learning is traditionally framed as a two-player zero-sum game between a learner and an adversarially chosen cost function. In recent years, diffusion models have emerged as a non-adversarial alternative to GANs. We show our approach outperforms GAN-style imitation learning baselines across various continuous control problems.
  arXiv  Detail & Related papers  (2024-10-17T17:59:25Z)
• Efficient Distribution Matching of Representations via Noise-Injected Deep InfoMax [73.03684002513218]
  We enhance Deep InfoMax (DIM) to enable automatic matching of learned representations to a selected prior distribution. We show that such modification allows for learning uniformly and normally distributed representations. The results indicate a moderate trade-off between the performance on the downstream tasks and quality of DM.
  arXiv  Detail & Related papers  (2024-10-09T15:40:04Z)
• Enhancing Knowledge Distillation of Large Language Models through Efficient Multi-Modal Distribution Alignment [10.104085497265004]
  We propose Ranking Loss based Knowledge Distillation (RLKD), which encourages consistency of peak predictions between the teacher and student models. Our method enables the student model to better learn the multi-modal distributions of the teacher model, leading to a significant performance improvement in various downstream tasks.
  arXiv  Detail & Related papers  (2024-09-19T08:06:42Z)
• MetaRM: Shifted Distributions Alignment via Meta-Learning [52.94381279744458]
  Reinforcement Learning from Human Feedback (RLHF) in language model alignment is critically dependent on the capability of the reward model (RM) We introduce MetaRM, a method leveraging meta-learning to align the RM with the shifted environment distribution. Extensive experiments demonstrate that MetaRM significantly improves the RM's distinguishing ability in iterative RLHF optimization.
  arXiv  Detail & Related papers  (2024-05-01T10:43:55Z)
• A Distributional Analogue to the Successor Representation [54.99439648059807]
  This paper contributes a new approach for distributional reinforcement learning. It elucidates a clean separation of transition structure and reward in the learning process. As an illustration, we show that it enables zero-shot risk-sensitive policy evaluation.
  arXiv  Detail & Related papers  (2024-02-13T15:35:24Z)
• Diff-Instruct: A Universal Approach for Transferring Knowledge From Pre-trained Diffusion Models [77.83923746319498]
  We propose a framework called Diff-Instruct to instruct the training of arbitrary generative models. We show that Diff-Instruct results in state-of-the-art single-step diffusion-based models. Experiments on refining GAN models show that the Diff-Instruct can consistently improve the pre-trained generators of GAN models.
  arXiv  Detail & Related papers  (2023-05-29T04:22:57Z)
• Distributionally Robust Models with Parametric Likelihood Ratios [123.05074253513935]
  Three simple ideas allow us to train models with DRO using a broader class of parametric likelihood ratios. We find that models trained with the resulting parametric adversaries are consistently more robust to subpopulation shifts when compared to other DRO approaches.
  arXiv  Detail & Related papers  (2022-04-13T12:43:12Z)
• A Distributional Approach to Controlled Text Generation [3.279201607581627]
  We propose a Distributional Approach to address Controlled Text Generation from pre-trained Language Models (LMs) This view permits to define, in a single formal framework, "pointwise" and "distributional" constraints over the target LM. We then perform experiments over distributional constraints, a unique feature of our approach, demonstrating its potential as a remedy to the problem of Bias in Language Models.
  arXiv  Detail & Related papers  (2020-12-21T19:02:41Z)
• Unsupervised Paraphrasing with Pretrained Language Models [85.03373221588707]
  We propose a training pipeline that enables pre-trained language models to generate high-quality paraphrases in an unsupervised setting. Our recipe consists of task-adaptation, self-supervision, and a novel decoding algorithm named Dynamic Blocking. We show with automatic and human evaluations that our approach achieves state-of-the-art performance on both the Quora Question Pair and the ParaNMT datasets.
  arXiv  Detail & Related papers  (2020-10-24T11:55:28Z)

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.