Skip to main content
Cornell University
We gratefully acknowledge support from the Simons Foundation, member institutions, and all contributors. Donate
arxiv logo > cs > arXiv:2411.04427

Help | Advanced Search

arXiv logo
Cornell University Logo

quick links

  • Login
  • Help Pages
  • About

Computer Science > Computation and Language

arXiv:2411.04427 (cs)
[Submitted on 7 Nov 2024 (v1), last revised 7 Jul 2025 (this version, v3)]

Title:One fish, two fish, but not the whole sea: Alignment reduces language models' conceptual diversity

Authors:Sonia K. Murthy, Tomer Ullman, Jennifer Hu
View a PDF of the paper titled One fish, two fish, but not the whole sea: Alignment reduces language models' conceptual diversity, by Sonia K. Murthy and 2 other authors
View PDF HTML (experimental)
Abstract:Researchers in social science and psychology have recently proposed using large language models (LLMs) as replacements for humans in behavioral research. In addition to arguments about whether LLMs accurately capture population-level patterns, this has raised questions about whether LLMs capture human-like conceptual diversity. Separately, it is debated whether post-training alignment (RLHF or RLAIF) affects models' internal diversity. Inspired by human studies, we use a new way of measuring the conceptual diversity of synthetically-generated LLM "populations" by relating the internal variability of simulated individuals to the population-level variability. We use this approach to evaluate non-aligned and aligned LLMs on two domains with rich human behavioral data. While no model reaches human-like diversity, aligned models generally display less diversity than their instruction fine-tuned counterparts. Our findings highlight potential trade-offs between increasing models' value alignment and decreasing the diversity of their conceptual representations.
Comments: 17 pages, 10 figures; updated with publishing information
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2411.04427 [cs.CL]
  (or arXiv:2411.04427v3 [cs.CL] for this version)
  https://doi.org/10.48550/arXiv.2411.04427
arXiv-issued DOI via DataCite
Journal reference: Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)
Related DOI: https://doi.org/10.18653/v1/2025.naacl-long.561
DOI(s) linking to related resources

Submission history

From: Sonia Murthy [view email]
[v1] Thu, 7 Nov 2024 04:38:58 UTC (4,667 KB)
[v2] Tue, 12 Nov 2024 20:11:58 UTC (4,664 KB)
[v3] Mon, 7 Jul 2025 18:34:06 UTC (4,655 KB)
Full-text links:

Access Paper:

    View a PDF of the paper titled One fish, two fish, but not the whole sea: Alignment reduces language models' conceptual diversity, by Sonia K. Murthy and 2 other authors
  • View PDF
  • HTML (experimental)
  • TeX Source
  • Other Formats
license icon view license
Current browse context:
cs.CL
< prev   |   next >
new | recent | 2024-11
Change to browse by:
cs

References & Citations

  • NASA ADS
  • Google Scholar
  • Semantic Scholar
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