Skip to main content
Cornell University
We gratefully acknowledge support from the Simons Foundation, member institutions, and all contributors. Donate
arxiv logo > cs > arXiv:2211.08411

Help | Advanced Search

arXiv logo
Cornell University Logo

quick links

  • Login
  • Help Pages
  • About

Computer Science > Computation and Language

arXiv:2211.08411 (cs)
[Submitted on 15 Nov 2022 (v1), last revised 27 Jul 2023 (this version, v2)]

Title:Large Language Models Struggle to Learn Long-Tail Knowledge

Authors:Nikhil Kandpal, Haikang Deng, Adam Roberts, Eric Wallace, Colin Raffel
View a PDF of the paper titled Large Language Models Struggle to Learn Long-Tail Knowledge, by Nikhil Kandpal and 4 other authors
View PDF
Abstract:The Internet contains a wealth of knowledge -- from the birthdays of historical figures to tutorials on how to code -- all of which may be learned by language models. However, while certain pieces of information are ubiquitous on the web, others appear extremely rarely. In this paper, we study the relationship between the knowledge memorized by large language models and the information in pre-training datasets scraped from the web. In particular, we show that a language model's ability to answer a fact-based question relates to how many documents associated with that question were seen during pre-training. We identify these relevant documents by entity linking pre-training datasets and counting documents that contain the same entities as a given question-answer pair. Our results demonstrate strong correlational and causal relationships between accuracy and relevant document count for numerous question answering datasets (e.g., TriviaQA), pre-training corpora (e.g., ROOTS), and model sizes (e.g., 176B parameters). Moreover, while larger models are better at learning long-tail knowledge, we estimate that today's models must be scaled by many orders of magnitude to reach competitive QA performance on questions with little support in the pre-training data. Finally, we show that retrieval-augmentation can reduce the dependence on relevant pre-training information, presenting a promising approach for capturing the long-tail.
Comments: ICML 2023 Camera Ready Version
Subjects: Computation and Language (cs.CL); Machine Learning (cs.LG)
Cite as: arXiv:2211.08411 [cs.CL]
  (or arXiv:2211.08411v2 [cs.CL] for this version)
  https://doi.org/10.48550/arXiv.2211.08411
arXiv-issued DOI via DataCite

Submission history

From: Nikhil Kandpal [view email]
[v1] Tue, 15 Nov 2022 18:49:27 UTC (5,479 KB)
[v2] Thu, 27 Jul 2023 08:01:42 UTC (5,503 KB)
Full-text links:

Access Paper:

    View a PDF of the paper titled Large Language Models Struggle to Learn Long-Tail Knowledge, by Nikhil Kandpal and 4 other authors
  • View PDF
  • TeX Source
  • Other Formats
license icon view license
Current browse context:
cs.CL
< prev   |   next >
new | recent | 2022-11
Change to browse by:
cs
cs.LG

References & Citations

  • NASA ADS
  • Google Scholar
  • Semantic Scholar

2 blog links

(what is this?)
a export BibTeX citation Loading...

BibTeX formatted citation

×
Data provided by:

Bookmark

BibSonomy logo Reddit logo

Bibliographic and Citation Tools

Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)

Code, Data and Media Associated with this Article

alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)

Demos

Replicate (What is Replicate?)
Hugging Face Spaces (What is Spaces?)
TXYZ.AI (What is TXYZ.AI?)

Recommenders and Search Tools

Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
  • Author
  • Venue
  • Institution
  • Topic

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)
  • About
  • Help
  • contact arXivClick here to contact arXiv Contact
  • subscribe to arXiv mailingsClick here to subscribe Subscribe
  • Copyright
  • Privacy Policy
  • Web Accessibility Assistance
  • arXiv Operational Status
    Get status notifications via email or slack