Memorizing Documents with Guidance in Large Language Models

Authors: Bumjin Park, Jaesik Choi

IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on Wikitext-103-v1 with Pythia1B show that the proposed methods provide different memory entries for documents and high recall of document-related content in generation with trained document-wise memories.
Researcher Affiliation Academia Bumjin Park 1 and Jaesik Choi1,2 1KAIST AI 2INEEJI {bumjin, jaesik.choi}@kaist.ac.kr
Pseudocode No The paper does not include a dedicated section or figure for pseudocode or algorithm blocks.
Open Source Code Yes We make the source code publicly available.2 https://github.com/fxnnxc/Doc Guidance LLM
Open Datasets Yes We train Pythia 1B [Biderman et al., 2023] to memorize Wikitext-103-v1 [Merity et al., 2017]
Dataset Splits No The paper mentions using Wikitext-103-v1 but does not explicitly provide details about specific train/validation/test dataset splits used for the experiments.
Hardware Specification No The paper does not provide specific hardware details such as GPU models, CPU types, or memory used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, CUDA versions).
Experiment Setup Yes We individually train document-wise memories for 10, 20, and 50 documents with guidance α = 0.1 and τ = 2.5. For baselines, we train two types of memory modules without guidance. Shared is the MLP in Equation 3, and Add is a module that directly adds differential memory entries. We also evaluate three activation types for document memory entries: Re LU, Tanh, and Sigmoid, which affect memory selections.