Tsinghua, Oxford and Stanford paper finds LLMs still write out reasoning with thinking disabled

Back Rohan Paul

Rohan Paul· @rohanpaul_ai · X· 2026-10-10 12:10 · Original publication time AI score54

Automatically verified and published · Generated and evidence-checked automatically; not reviewed by a human.

AI introduction

The paper reports that, with thinking disabled, DeepSeek-V4-Flash still wrote out reasoning in 99.9% of open-ended answers. Forcing answer-only replies on open-ended tasks raised compliance to about 40% across five models, but reduced accuracy by about 15 percentage points.

Article · Original

The article text is unavailable in this language; an existing version is shown.

New Tsinghua, Oxford, and Stanford paper finds that LLMs keep reasoning out loud even with thinking turned off, especially on open-ended questions.

With thinking disabled, DeepSeek-V4-Flash still wrote out reasoning in 99.9% of open-ended answers. A missing think tag or a short reply does not prove the model skipped reasoning.

Yes/no questions are easy to answer directly, multiple-choice questions sit in the middle, and open-ended ones keep pulling the model back into reasoning.

Forcing answer-only replies on open-ended tasks raised compliance to about 40% across 5 models but cut accuracy by about 15 points.

– arxiv. org/abs/2610.11765

Title: "Thinking Inertia: LLMs Keep Thinking When Told Not To"

Found an error? Send a correction