paper-with-me

Papers

KILM: Knowledge Injection into Encoder-Decoder Language Models

2023-02-17 · Yan Xu, Mahdi Namazifar, Devamanyu Hazarika, Aishwarya Padmakumar, Yang Liu, Dilek Hakkani-Tür

Large pre-trained language models (PLMs) have been shown to retain implicit knowledge within their parameters. To enhance this implicit knowledge, we propose Knowledge Injection into Language Models (KILM), a novel approach that injects entity-related knowledge into encoder-decoder PLMs, via a generative knowledge infilling objective through continued pre-training. This is done without architectural modifications to the PLMs or adding additional parameters. Experimental results over a suite of knowledge-intensive tasks spanning numerous datasets show that KILM enables models to retain more knowledge and hallucinate less, while preserving their original performance on general NLU and NLG tasks. KILM also demonstrates improved zero-shot performances on tasks such as entity disambiguation, outperforming state-of-the-art models having 30x more parameters.

📄 PDF Abstract BibTeX arXiv:2302.09170

Code (1)

alexa/kilm 공식 구현 pytorch

Tasks

DecoderEntity Disambiguation

Similar Papers 제목 키워드 기반

LokiLM: Technical Report

2024-07-10 · Justin Kiefel, Shrey Shah

In this work, we introduce LokiLM, a 1.4B parameter large language model trained on 500B tokens. Our model performs strongly in natural language reasoning tasks and achieves state-of-the-art performance among models with…

Knowledge DistillationLanguage ModelingLanguage ModellingLarge Language Model+1

LaKo: Knowledge-driven Visual Question Answering via Late Knowledge-to-Text Injection

2022-07-26 · Zhuo Chen, Yufeng Huang, Jiaoyan Chen, Yuxia Geng 외

Visual question answering (VQA) often requires an understanding of visual concepts and language semantics, which relies on external knowledge. Most existing methods exploit pre-trained language models or/and unstructured…

DecoderKnowledge GraphsQuestion AnsweringText Generation+2

An Enhanced Knowledge Injection Model for Commonsense Generation

2020-12-01 · COLING 2020 8 · Zhihao Fan, Yeyun Gong, Zhongyu Wei, Siyuan Wang 외

Commonsense generation aims at generating plausible everyday scenario description based on a set of provided concepts. Digging the relationship of concepts from scratch is non-trivial, therefore, we retrieve prototypes f…

DecodermodelPosition

Languages are Modalities: Cross-Lingual Alignment via Encoder Injection

2025-10-31 · Rajan Agarwal, Aarush Gupta arxiv

Instruction-tuned Large Language Models (LLMs) underperform on low resource, non-Latin scripts due to tokenizer fragmentation and weak cross-lingual coupling. We present LLINK (Latent Language Injection for Non-English K…

UNTER: A Unified Knowledge Interface for Enhancing Pre-trained Language Models

2023-05-02 · Deming Ye, Yankai Lin, Zhengyan Zhang, Maosong Sun

Recent research demonstrates that external knowledge injection can advance pre-trained language models (PLMs) in a variety of downstream NLP tasks. However, existing knowledge injection methods are either applicable to s…

DecoderEntity Typingnamed-entity-recognitionNamed Entity Recognition+1