DAFNet: Dynamic Auxiliary Fusion for Sequential Model Editing in Large Language Models

• URL: http://arxiv.org/abs/2405.20588v1
• Date: Fri, 31 May 2024 02:56:49 GMT
• Title: DAFNet: Dynamic Auxiliary Fusion for Sequential Model Editing in Large Language Models
• Authors: Taolin Zhang, Qizhou Chen, Dongyang Li, Chengyu Wang, Xiaofeng He, Longtao Huang, Hui Xue, Jun Huang,
• Abstract summary: A Dynamic Auxiliary Fusion Network (DAFNet) is designed to enhance the semantic interaction among the factual knowledge within the entire sequence. DAFNet significantly outperforms strong baselines in single-turn and sequential editing.
• Score: 32.598670876662375
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Recently, while large language models (LLMs) have demonstrated impressive results, they still suffer from hallucination, i.e., the generation of false information. Model editing is the task of fixing factual mistakes in LLMs; yet, most previous works treat it as a one-time task, paying little attention to ever-emerging mistakes generated by LLMs. We address the task of sequential model editing (SME) that aims to rectify mistakes continuously. A Dynamic Auxiliary Fusion Network (DAFNet) is designed to enhance the semantic interaction among the factual knowledge within the entire sequence, preventing catastrophic forgetting during the editing process of multiple knowledge triples. Specifically, (1) for semantic fusion within a relation triple, we aggregate the intra-editing attention flow into auto-regressive self-attention with token-level granularity in LLMs. We further leverage multi-layer diagonal inter-editing attention flow to update the weighted representations of the entire sequence-level granularity. (2) Considering that auxiliary parameters are required to store the knowledge for sequential editing, we construct a new dataset named \textbf{DAFSet}, fulfilling recent, popular, long-tail and robust properties to enhance the generality of sequential editing. Experiments show DAFNet significantly outperforms strong baselines in single-turn and sequential editing. The usage of DAFSet also consistently improves the performance of other auxiliary network-based methods in various scenarios

Related papers

• LiNeS: Post-training Layer Scaling Prevents Forgetting and Enhances Model Merging [80.17238673443127]
  LiNeS is a post-training editing technique designed to preserve pre-trained generalization while enhancing fine-tuned task performance. LiNeS demonstrates significant improvements in both single-task and multi-task settings across various benchmarks in vision and natural language processing.
  arXiv  Detail & Related papers  (2024-10-22T16:26:05Z)
• Neuron-Level Sequential Editing for Large Language Models [19.324852774144752]
  We introduce textbfNeuron-level textbfSequential textbfEditing (NSE) for supporting sequential model editing. Specifically, we optimize the target layer's hidden states using the model's original weights to prevent model failure. Our experiments demonstrate that NSE significantly outperforms current modifying parameters model editing methods.
  arXiv  Detail & Related papers  (2024-10-05T05:52:22Z)
• Enhance Lifelong Model Editing with Continuous Data-Adapter Association [55.697627106315004]
  Large language models (LLMs) require model editing to efficiently update specific knowledge within them and avoid factual errors. Current approaches manage sequential edits by freezing original parameters and allocating new adapters for each knowledge modification. We propose ELDER, textbfEnhancing textbfLifelong motextbfDel textbfEditing with mixtutextbfRe of Low-Rank Adapter (LoRA)
  arXiv  Detail & Related papers  (2024-08-19T02:27:00Z)
• Perturbation-Restrained Sequential Model Editing [33.51709226068619]
  Current model editing methods compromise the general abilities of large language models (LLMs) as the number of edits increases. We propose a framework termed Perturbation Restraint on Upper bouNd for Editing (PRUNE) PRUNE can preserve considerable general abilities while maintaining the editing performance effectively in sequential model editing.
  arXiv  Detail & Related papers  (2024-05-27T04:40:56Z)
• The Butterfly Effect of Model Editing: Few Edits Can Trigger Large Language Models Collapse [58.0132400208411]
  Even a single edit can trigger model collapse, manifesting as significant performance degradation in various benchmark tasks. benchmarking Large Language Models after each edit is impractically time-consuming and resource-intensive. We have utilized GPT-3.5 to develop a new dataset, HardEdit, based on hard cases.
  arXiv  Detail & Related papers  (2024-02-15T01:50:38Z)
• Massive Editing for Large Language Models via Meta Learning [27.972194696587813]
  Large language models (LLMs) have enabled learning knowledge from the pre-training corpora, but the acquired knowledge may be fundamentally incorrect or outdated over time. We propose the MAssive Language Model Editing Network (MALMEN), which formulates the parameter shift aggregation as the least square problem. Our method is evaluated by editing up to thousands of facts on LMs with different architectures, i.e., BERT-base, GPT-2, T5-XL (2.8B), and GPT-J (6B)
  arXiv  Detail & Related papers  (2023-11-08T13:03:06Z)
• AdaMerging: Adaptive Model Merging for Multi-Task Learning [68.75885518081357]
  This paper introduces an innovative technique called Adaptive Model Merging (AdaMerging) It aims to autonomously learn the coefficients for model merging, either in a task-wise or layer-wise manner, without relying on the original training data. Compared to the current state-of-the-art task arithmetic merging scheme, AdaMerging showcases a remarkable 11% improvement in performance.
  arXiv  Detail & Related papers  (2023-10-04T04:26:33Z)
• SequenceMatch: Imitation Learning for Autoregressive Sequence Modelling with Backtracking [60.109453252858806]
  A maximum-likelihood (MLE) objective does not match a downstream use-case of autoregressively generating high-quality sequences. We formulate sequence generation as an imitation learning (IL) problem. This allows us to minimize a variety of divergences between the distribution of sequences generated by an autoregressive model and sequences from a dataset. Our resulting method, SequenceMatch, can be implemented without adversarial training or architectural changes.
  arXiv  Detail & Related papers  (2023-06-08T17:59:58Z)
• Memory-Based Model Editing at Scale [102.28475739907498]
  Existing model editors struggle to accurately model an edit's intended scope. We propose Semi-Parametric Editing with a Retrieval-Augmented Counterfactual Model (SERAC) SERAC stores edits in an explicit memory and learns to reason over them to modulate the base model's predictions as needed.
  arXiv  Detail & Related papers  (2022-06-13T23:40:34Z)
• FiD-Ex: Improving Sequence-to-Sequence Models for Extractive Rationale Generation [19.73842483996047]
  We develop FiD-Ex, which addresses shortcomings for seq2seq models by introducing sentence markers to eliminate explanation fabrication. FiD-Ex significantly improves over prior work in terms of explanation metrics and task accuracy, on multiple tasks from the ERASER explainability benchmark.
  arXiv  Detail & Related papers  (2020-12-31T07:22:15Z)

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.