Knowledge Tree
multi-task learning
Also known as: Multitask Learning, multi-task representation learning
Facts (16)
Sources
Practices, opportunities and challenges in the fusion of knowledge ... frontiersin.org 5 facts
referenceThe LP-BERT model (Li et al., 2023) employs a multi-task learning approach that simultaneously learns contextual and semantic information by sharing an input format across three tasks: Masked Language Model (MLM), Masked Entity Model (MEM), and Masked Relation Model (MRM).
referenceThe MT-DNN architecture proposed by Choi and Ko (2023) utilizes multi-task learning by combining Entity Description Prediction (EDP) and Entity Type Prediction (ITP) tasks, sharing a pre-trained language model and network layers for joint training.
referenceB. Kim, T. Hong, Y. Ko, and J. Seo published 'Multi-task learning for knowledge graph completion with pre-trained language models' in the Proceedings of the 28th International Conference on Computational Linguistics in 2020.
claimMulti-task learning approaches for knowledge graph completion, such as MT-DNN and LP-BERT, fail to resolve the fundamental scalability gap in large-scale knowledge graphs, where latency grows polynomially with graph density.
claimLi et al. (2021) introduced a breadth-first search (BFS) strategy with a relationship bias for knowledge graph linearization and employed multi-task learning with knowledge graph reconstruction.
Track: Poster Session 3 - aistats 2026 virtual.aistats.org 5 facts
claimMulti-task representation learning is widely used in deep learning applications, including computer vision and natural language processing, due to its generalization performance.
claimExperimental results across diverse healthcare datasets demonstrate that Adaptive Parameter Optimisation (APO) outperforms traditional information-sharing approaches, such as multi-task learning and model-agnostic meta-learning, in improving task performance.
procedureThe Collaborative Two-Sample Testing (CTST) framework integrates elements from f-divergence estimation, Kernel Methods, and Multitask Learning to efficiently leverage graph structure for two-sample testing.
claimMulti-task representation learning outperforms single-task representation learning in scenarios involving over-parameterized two-layer convolutional neural networks trained by gradient descent.
claimThe understanding of the underlying mechanisms of multi-task representation learning remains limited despite its widespread use.
Unlocking the Potential of Generative AI through Neuro-Symbolic ... arxiv.org Feb 16, 2025 2 facts
claimMulti-task learning in Neuro-Symbolic AI (NSAI) systems is defined as the capability to handle multiple tasks simultaneously through shared knowledge representations.
claimNeuro Symbolic Neuro, NeuroSymbolicLoss, and NeuroSymbolicNeuro architectures excel in multi-task learning and multi-domain adaptation, enabling effective reuse of knowledge across domains.
A survey on augmenting knowledge graphs (KGs) with large ... link.springer.com Nov 4, 2024 1 fact
claimLarge Language Models (LLMs) perform common sense reasoning and multi-task learning, allowing different tasks to be addressed within a single model.
Construction of Knowledge Graphs: State and Challenges - arXiv arxiv.org 1 fact
procedureThe AutoKnow system performs data imputation by extracting attribute-value pairs from product data using a taxonomy-aware tagging approach that leverages Conditional Random Fields (CRF) combined with multi-task learning and a shared BiLSTM to train sequence tagging and product type categorization.
Large Language Models Meet Knowledge Graphs for Question ... arxiv.org Sep 22, 2025 1 fact
referenceRAMQA enhances multi-modal retrieval-augmented question-answering by integrating learning-to-rank with the training of generative models via multi-task learning.
A Survey on the Theory and Mechanism of Large Language Models arxiv.org Mar 12, 2026 1 fact
claimZakerinia et al. (2025) propose that the strong generalization of highly overparameterized deep models can be explained by low intrinsic dimensionality from a multi-task learning perspective, where the learning process is confined to a low-dimensional manifold despite the vast number of parameters.