Instruction Tuned Models are Quick Learners Article Swipe

MAKE IMPACT

Himanshu Gupta , Saurabh Arjun Sawant , Swaroop Mishra , Mutsumi Nakamura , Arindam Mitra , Santosh Mashetty , Chitta Baral · YOU? · · 2023 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2306.05539

Instruction tuning of language models has demonstrated the ability to enhance model generalization to unseen tasks via in-context learning using a few examples. However, typical supervised learning still requires a plethora of downstream training data for finetuning. Often in real-world situations, there is a scarcity of data available for finetuning, falling somewhere between few shot inference and fully supervised finetuning. In this work, we demonstrate the sample efficiency of instruction tuned models over various tasks by estimating the minimal downstream training data required by them to perform transfer learning and match the performance of state-of-the-art (SOTA) supervised models. We conduct experiments on 119 tasks from Super Natural Instructions (SuperNI) in both the single task learning (STL) and multi task learning (MTL) settings. Our findings reveal that, in the STL setting, instruction tuned models equipped with 25% of the downstream train data surpass the SOTA performance on the downstream tasks. In the MTL setting, an instruction tuned model trained on only 6% of downstream training data achieve SOTA, while using 100% of the training data results in a 3.69% points improvement (ROUGE-L 74.68) over the previous SOTA. We conduct an analysis on T5 vs Tk-Instruct by developing several baselines to demonstrate that instruction tuning aids in increasing both sample efficiency and transfer learning. Additionally, we observe a consistent ~4% performance increase in both settings when pre-finetuning is performed with instructions. Finally, we conduct a categorical study and find that contrary to previous results, tasks in the question rewriting and title generation categories suffer from instruction tuning.

Type

preprint

Language

en

Landing Page

http://arxiv.org/abs/2306.05539

PDF

https://arxiv.org/pdf/2306.05539

OA Status

green

Cited By

3

Related Works

10

OpenAlex ID

https://openalex.org/W4380352309

Raw OpenAlex JSON

OpenAlex ID

https://openalex.org/W4380352309

Canonical identifier for this work in OpenAlex

DOI

https://doi.org/10.48550/arxiv.2306.05539

Digital Object Identifier

Title

Instruction Tuned Models are Quick Learners

Work title

Type

preprint

OpenAlex work type

Language

en

Primary language

Publication year

2023

Year of publication

Publication date

2023-05-17

Full publication date if available

Authors

Himanshu Gupta, Saurabh Arjun Sawant, Swaroop Mishra, Mutsumi Nakamura, Arindam Mitra, Santosh Mashetty, Chitta Baral

List of authors in order

Landing page

https://arxiv.org/abs/2306.05539

Publisher landing page

PDF URL

https://arxiv.org/pdf/2306.05539

Direct link to full text PDF

Open access

Yes

Whether a free full text is available

OA status

green

Open access status per OpenAlex

OA URL

https://arxiv.org/pdf/2306.05539

Direct OA link when available

Concepts

Computer science, Inference, Machine learning, Artificial intelligence, Transfer of learning, Generalization, Task (project management), Context (archaeology), Leverage (statistics), Labeled data, Mathematics, Biology, Management, Mathematical analysis, Paleontology, Economics

Top concepts (fields/topics) attached by OpenAlex

Cited by

3

Total citation count in OpenAlex

Citations by year (recent)

2024: 2, 2023: 1

Per-year citation counts (last 5 years)

Related works (count)

10

Other works algorithmically related by OpenAlex

Full payload