Encoder Fine-tuning with Stochastic Sampling Outperforms Open-weight GPT in Astronomy Knowledge Extraction

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Rawat, Shivam, Flek, Lucie, Karimi, Akbar
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917073191436288
author Rawat, Shivam
Flek, Lucie
Karimi, Akbar
author_facet Rawat, Shivam
Flek, Lucie
Karimi, Akbar
contents Scientific literature in astronomy is rapidly expanding, making it increasingly important to automate the extraction of key entities and contextual information from research papers. In this paper, we present an encoder-based system for extracting knowledge from astronomy articles. Our objective is to develop models capable of classifying telescope references, detecting auxiliary semantic attributes, and recognizing instrument mentions from textual content. To this end, we implement a multi-task transformer-based system built upon the SciBERT model and fine-tuned for astronomy corpora classification. To carry out the fine-tuning, we stochastically sample segments from the training data and use majority voting over the test segments at inference time. Our system, despite its simplicity and low-cost implementation, significantly outperforms the open-weight GPT baseline.
format Preprint
id arxiv_https___arxiv_org_abs_2511_08204
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Encoder Fine-tuning with Stochastic Sampling Outperforms Open-weight GPT in Astronomy Knowledge Extraction
Rawat, Shivam
Flek, Lucie
Karimi, Akbar
Computation and Language
Scientific literature in astronomy is rapidly expanding, making it increasingly important to automate the extraction of key entities and contextual information from research papers. In this paper, we present an encoder-based system for extracting knowledge from astronomy articles. Our objective is to develop models capable of classifying telescope references, detecting auxiliary semantic attributes, and recognizing instrument mentions from textual content. To this end, we implement a multi-task transformer-based system built upon the SciBERT model and fine-tuned for astronomy corpora classification. To carry out the fine-tuning, we stochastically sample segments from the training data and use majority voting over the test segments at inference time. Our system, despite its simplicity and low-cost implementation, significantly outperforms the open-weight GPT baseline.
title Encoder Fine-tuning with Stochastic Sampling Outperforms Open-weight GPT in Astronomy Knowledge Extraction
topic Computation and Language
url https://arxiv.org/abs/2511.08204