Knowledge Editing in Language Models via Adapted Direct Preference Optimization

• URL: http://arxiv.org/abs/2406.09920v1
• Date: Fri, 14 Jun 2024 11:02:21 GMT
• Title: Knowledge Editing in Language Models via Adapted Direct Preference Optimization
• Authors: Amit Rozner, Barak Battash, Lior Wolf, Ofir Lindenbaum,
• Abstract summary: Large Language Models (LLMs) can become outdated over time. Knowledge Editing aims to overcome this challenge using weight updates that do not require expensive retraining.
• Score: 50.616875565173274
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large Language Models (LLMs) can become outdated over time as they may lack updated world knowledge, leading to factual knowledge errors and gaps. Knowledge Editing (KE) aims to overcome this challenge using weight updates that do not require expensive retraining. We propose treating KE as an LLM alignment problem. Toward this goal, we introduce Knowledge Direct Preference Optimization (KDPO), a variation of the Direct Preference Optimization (DPO) that is more effective for knowledge modifications. Our method is based on an online approach that continually updates the knowledge stored in the model. We use the current knowledge as a negative sample and the new knowledge we want to introduce as a positive sample in a process called DPO. We also use teacher-forcing for negative sample generation and optimize using the positive sample, which helps maintain localized changes. We tested our KE method on various datasets and models, comparing it to several cutting-edge methods, with 100 and 500 sequential edits. Additionally, we conducted an ablation study comparing our method to the standard DPO approach. Our experimental results show that our modified DPO method allows for more refined KE, achieving similar or better performance compared to previous methods.

Related papers

• Self-Augmented Preference Optimization: Off-Policy Paradigms for Language Model Alignment [104.18002641195442]
  We introduce Self-Augmented Preference Optimization (SAPO), an effective and scalable training paradigm that does not require existing paired data. Building on the self-play concept, which autonomously generates negative responses, we further incorporate an off-policy learning pipeline to enhance data exploration and exploitation.
  arXiv  Detail & Related papers  (2024-05-31T14:21:04Z)
• Mining the Explainability and Generalization: Fact Verification Based on Self-Instruction [0.7673339435080445]
  We propose a self-instruction based fine-tuning approach for fact-checking that balances accuracy and explainability. We fine-tune the smallest-scale LLaMA-7B model and evaluate it on the challenging fact-checking datasets FEVEROUS and HOVER. Our method is the first to leverage self-supervised learning for fact-checking and innovatively combines contrastive learning and improved DPO in fine-tuning LLMs.
  arXiv  Detail & Related papers  (2024-05-21T08:23:54Z)
• Monte Carlo Tree Search Boosts Reasoning via Iterative Preference Learning [55.96599486604344]
  We introduce an approach aimed at enhancing the reasoning capabilities of Large Language Models (LLMs) through an iterative preference learning process. We use Monte Carlo Tree Search (MCTS) to iteratively collect preference data, utilizing its look-ahead ability to break down instance-level rewards into more granular step-level signals. The proposed algorithm employs Direct Preference Optimization (DPO) to update the LLM policy using this newly generated step-level preference data.
  arXiv  Detail & Related papers  (2024-05-01T11:10:24Z)
• Unleashing the Potential of Large Language Models as Prompt Optimizers: An Analogical Analysis with Gradient-based Model Optimizers [108.72225067368592]
  We propose a novel perspective to investigate the design of large language models (LLMs)-based prompts. We identify two pivotal factors in model parameter learning: update direction and update method. In particular, we borrow the theoretical framework and learning methods from gradient-based optimization to design improved strategies.
  arXiv  Detail & Related papers  (2024-02-27T15:05:32Z)
• Active Preference Learning for Large Language Models [12.093302163058436]
  We develop an active learning strategy for DPO to make better use of preference labels. We propose a practical acquisition function for prompt/completion pairs based on the predictive entropy of the language model. We demonstrate how our approach improves both the rate of learning and final performance of fine-tuning on pairwise preference data.
  arXiv  Detail & Related papers  (2024-02-12T23:09:00Z)
• Model-based adaptation for sample efficient transfer in reinforcement learning control of parameter-varying systems [1.8799681615947088]
  We leverage ideas from model-based control to address the sample efficiency problem of reinforcement learning algorithms. We demonstrate that our approach is more sample-efficient than fine-tuning with reinforcement learning alone.
  arXiv  Detail & Related papers  (2023-05-20T10:11:09Z)
• Weighted Ensemble Self-Supervised Learning [67.24482854208783]
  Ensembling has proven to be a powerful technique for boosting model performance. We develop a framework that permits data-dependent weighted cross-entropy losses. Our method outperforms both in multiple evaluation metrics on ImageNet-1K.
  arXiv  Detail & Related papers  (2022-11-18T02:00:17Z)
• Towards Evaluating Adaptivity of Model-Based Reinforcement Learning Methods [25.05409184943328]
  We show that well-known model-based methods perform poorly in their ability to adapt to local environmental changes. We identify elements that hurt adaptive behavior and link these to underlying techniques frequently used in deep model-based RL. We provide insights into the challenges of building an adaptive nonlinear model-based method.
  arXiv  Detail & Related papers  (2022-04-25T06:45:16Z)
• Few-shot Quality-Diversity Optimization [50.337225556491774]
  Quality-Diversity (QD) optimization has been shown to be effective tools in dealing with deceptive minima and sparse rewards in Reinforcement Learning. We show that, given examples from a task distribution, information about the paths taken by optimization in parameter space can be leveraged to build a prior population, which when used to initialize QD methods in unseen environments, allows for few-shot adaptation. Experiments carried in both sparse and dense reward settings using robotic manipulation and navigation benchmarks show that it considerably reduces the number of generations that are required for QD optimization in these environments.
  arXiv  Detail & Related papers  (2021-09-14T17:12:20Z)
• META-Learning Eligibility Traces for More Sample Efficient Temporal Difference Learning [2.0559497209595823]
  We propose a meta-learning method for adjusting the eligibility trace parameter, in a state-dependent manner. The adaptation is achieved with the help of auxiliary learners that learn distributional information about the update targets online. We prove that, under some assumptions, the proposed method improves the overall quality of the update targets, by minimizing the overall target error.
  arXiv  Detail & Related papers  (2020-06-16T03:41:07Z)

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.