Stop Anthropomorphizing Intermediate Tokens as Reasoning/Thinking Traces (2025)
Mirrored from Hacker News — AI on Front Page for archival readability. Support the source by reading on the original site.
Computer Science > Artificial Intelligence
arXiv:2504.09762 (cs)
[Submitted on 14 Apr 2025 (v1), last revised 9 Jun 2026 (this version, v4)]
Title:Position: Stop Anthropomorphizing Intermediate Tokens as Reasoning/Thinking Traces!
Authors:Subbarao Kambhampati, Karthik Valmeekam, Siddhant Bhambri, Vardhan Palod, Lucas Saldyt, Kaya Stechly, Soumya Rani Samineni, Durgesh Kalwar, Upasana Biswas
View a PDF of the paper titled Position: Stop Anthropomorphizing Intermediate Tokens as Reasoning/Thinking Traces!, by Subbarao Kambhampati and 8 other authors
View PDF
HTML (experimental)
Abstract:Intermediate token generation (ITG), where a model produces output before the solution, has become a standard method to improve the performance of language models on reasoning tasks. These intermediate tokens have been called \say{reasoning traces} or even \say{thinking traces} -- implicitly anthropomorphizing the traces, and implying that these traces resemble steps a human might take when solving a challenging problem, and as such can provide an interpretable window into the operation of the model's thinking process to the end user. In this position paper, we present evidence that this anthropomorphization isn't a harmless metaphor, and instead is quite dangerous -- it confuses the nature of these models and how to use them effectively, and leads to questionable research. We call on the community to avoid such anthropomorphization of intermediate tokens.
| Comments: | Appears in ICML 2026. [This is a fork of v1. This fork, while overlapping with v1 in background section, differs both in the overall focus as well as the specific argument against anthropomorphization of reasoning traces] |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2504.09762 [cs.AI] |
| (or arXiv:2504.09762v4 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2504.09762
arXiv-issued DOI via DataCite
|
|
| Journal reference: | ICML 2026 |
Submission history
From: Subbarao Kambhampati [view email][v1] Mon, 14 Apr 2025 00:03:34 UTC (381 KB)
[v2] Tue, 27 May 2025 16:35:47 UTC (536 KB)
[v3] Fri, 6 Mar 2026 14:36:07 UTC (1,321 KB)
[v4] Tue, 9 Jun 2026 20:08:39 UTC (4,278 KB)
Full-text links:
Access Paper:
- View PDF
- HTML (experimental)
- TeX Source
View a PDF of the paper titled Position: Stop Anthropomorphizing Intermediate Tokens as Reasoning/Thinking Traces!, by Subbarao Kambhampati and 8 other authors
References & Citations
Loading...
Bookmark
Bibliographic Tools
Code, Data, Media
Demos
Related Papers
About arXivLabs
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 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
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
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.