SciRIFF: A Resource to Enhance Language Model Instruction-Following over Scientific Literature

TL;DR

SciRIFF is a high-quality, expert-written dataset of 137K instruction-following instances across scientific tasks, enabling improved language model performance in scientific literature understanding.

Contribution

The paper introduces SciRIFF, a novel dataset designed to enhance instruction-following capabilities of language models specifically for scientific literature tasks.

Findings

LLMs finetuned on SciRIFF outperform baselines by 70.6% on out-of-distribution tasks.

SciRIFF covers 54 scientific tasks with complex instructions and structured outputs.

The dataset improves LLMs' ability to extract, summarize, and verify scientific information.

Abstract

We present SciRIFF (Scientific Resource for Instruction-Following and Finetuning), a dataset of 137K instruction-following instances for training and evaluation, covering 54 tasks. These tasks span five core scientific literature understanding capabilities: information extraction, summarization, question answering, claim verification, and classification. SciRIFF is unique in being entirely expert-written, high-quality instruction-following dataset for extracting and synthesizing information from research literature across diverse scientific fields. It features complex instructions with long input contexts, detailed task descriptions, and structured outputs. To demonstrate its utility, we finetune a series of large language models (LLMs) using a mix of general-domain and SciRIFF instructions. On nine out-of-distribution held-out tasks (referred to as SciRIFF-Eval), LLMs finetuned on SciRIFF achieve 70.6% average improvement over baselines trained only on general-domain instructions. SciRIFF facilitates the development and evaluation of LLMs to help researchers navigate the rapidly growing body of scientific literature.