Skip to content
AI News HubLIVE
Original source2 min read

Conserved Verdicts, Drifting Commitments: Identifying Harmful Overthinking

Summary

An arXiv paper introduces a commitment-versus-verdict framework that separates the correctness of a model's current answer from the correctness of its final answer, shows that final-correctness probes alone cannot decide when to stop reasoning, and finds that continuing to think costs an average of 3.6 points at points where DeepSeek-R1-Distill-Qwen-1.5B already holds a correct answer on MATH-500.

SourcearXiv AIAuthor: Jianghong Zhou, Hao Rong, Chen Lin
Conserved Verdicts, Drifting Commitments: Identifying Harmful Overthinking
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 16 Nov 2026]

Title:Conserved Verdicts, Drifting Commitments: Identifying Harmful Overthinking

View a PDF of the paper titled Conserved Verdicts, Drifting Commitments: Identifying Harmful Overthinking, by Jianghong Zhou and 2 other authors

View PDF HTML (experimental)

Abstract:A reasoning model that has reached a correct answer can lose it by continuing to think. Deciding whether to stop means comparing two probabilities at the same point of a chain: that the answer the model would give if stopped now is correct (its commitment), and that the answer it will finish with is correct (its verdict). We show that the verdict alone, which is what final-correctness probes estimate, cannot make this comparison. As a forecast of a fixed outcome it has no predictable drift, whether continuing helps or hurts. The commitment-verdict gap equals the expected remaining change in answer quality and the accuracy gained by stopping. A second identity splits any accuracy-versus-budget curve into answer change within chains and change in which chains are still running. We measure both on MATH-500. At all 379 points where DeepSeek-R1-Distill-Qwen-1.5B holds a correct answer, selected without regard to how its chain ends, continuing loses 3.6 points on average (95% CI [2.2, 5.2]), and the loss concentrates in a few problems. In six reasoning models from three families, on MATH-500 and OlympiadBench and with budgets up to 16,384 tokens, declining accuracy curves come from which chains remain, while the chains still reasoning keep or improve their answers. An oracle stopping rule gains at most +1.8 and +2.5 points at 1.5B and 7B. The framework says what to measure before building a stopping rule.

Comments: 40 pages

Subjects:

Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)

Cite as: arXiv:2611.21946 [cs.AI]

(or arXiv:2611.21946v1 [cs.AI] for this version)

https://doi.org/10.48550/arXiv.2611.21946

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Jianghong Zhou [view email] [v1] Mon, 16 Nov 2026 22:08:54 UTC (633 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Conserved Verdicts, Drifting Commitments: Identifying Harmful Overthinking, by Jianghong Zhou and 2 other authors

View PDF

HTML (experimental)

TeX Source

view license

Additional Features

Audio Summary

Current browse context:

cs.AI

new | recent | 2026-11

Change to browse by:

cs cs.CL cs.LG

References & Citations

SciX

Google Scholar

Semantic Scholar

Loading...

Data provided by:

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?)

Author

Venue

Institution

Topic

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.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • Commitment is the probability that the answer a model would give if stopped now is correct; verdict is the probability that the answer it will finish with is correct, and their gap equals the expected remaining change in answer quality and the accuracy gained by stopping.
  • Final-correctness probes estimate only the verdict and cannot predict whether continuing to think will help or hurt.
  • Across all 379 points where DeepSeek-R1-Distill-Qwen-1.5B already holds a correct answer, continuing loses 3.6 points on average (95% CI [2.2, 5.2]), concentrated in a few problems.
  • In six reasoning models from three families on MATH-500 and OlympiadBench with budgets up to 16,384 tokens, declining accuracy curves come from which chains remain, while chains still reasoning keep or improve their answers.
  • An oracle stopping rule gains at most +1.8 and +2.5 points at 1.5B and 7B, so the framework says what to measure before building a stopping rule.

Highlights and analysis are generated automatically and may contain errors. Check the original source.

Conserved Verdicts, Drifting Commitments: Identifying Harmful Overthinking | AI News Hub