Prismix
arXiv — NLP / Computation & Language · · 3 min read

Language Models Compare Quantities Using Number-specific and Unit-specific Heuristics

#paper

Mirrored from arXiv — NLP / Computation & Language for archival readability. Support the source by reading on the original site.

Like Read original ↗

Computer Science > Computation and Language

arXiv:2606.03982 (cs)
[Submitted on 2 Jun 2026]

Title:Language Models Compare Quantities Using Number-specific and Unit-specific Heuristics

View a PDF of the paper titled Language Models Compare Quantities Using Number-specific and Unit-specific Heuristics, by Mutsumi Sasaki and 6 other authors
View PDF
Abstract:Quantities with measurement units, such as 110 cm and 1.2 m, require language models (LMs) to combine a numeral with a symbolic unit scale. Here, we study how LMs compare such quantities in controlled settings spanning several unit systems. We find that accuracy degrades near the comparison boundary, where small changes in value determine the correct answer. The resulting errors are systematic: linear surrogate models predict LM preferences from numerical-difference and unit-scale-difference cues, and causal interventions on subspaces aligned with these variables shift model's output. The results suggest that LMs compare quantities through a bag of heuristics over numerals and units, rather than first converting both expressions to an exact shared-scale representation.
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2606.03982 [cs.CL]
  (or arXiv:2606.03982v1 [cs.CL] for this version)
  https://doi.org/10.48550/arXiv.2606.03982
arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Mutsumi Sasaki [view email]
[v1] Tue, 2 Jun 2026 17:58:02 UTC (3,198 KB)
Full-text links:

Access Paper:

    View a PDF of the paper titled Language Models Compare Quantities Using Number-specific and Unit-specific Heuristics, by Mutsumi Sasaki and 6 other authors
  • View PDF
  • TeX Source

Current browse context:

cs.CL
< prev   |   next >
Change to browse by:
cs

References & Citations

Loading...

BibTeX formatted citation

loading...
Data provided by:

Bookmark

BibSonomy Reddit
Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media

Code, Data and Media Associated with this Article

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

Demos

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

Recommenders and Search Tools

Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
About arXivLabs

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.

Discussion (0)

Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.

Sign in →

No comments yet. Sign in and be the first to say something.

More from arXiv — NLP / Computation & Language