Claim◆Article

The <think></think> tags are purely cosmetic markers for separating the reasoning trace from the final answer; they do not confer or improve reasoning ability, and a model trained without them would likely perform just as well.

Raschka argues that think tokens are cosmetic formatting devices, not the source of a reasoning model's ability to reason. ✦ AI generated

Sebastian Raschka · Ahead of AI · 2026-07-18 · original ↗

Share card ↓Export .md Export .txt

These <think> and </think> tags are cosmetic with respect to reasoning ability. They do not make the model reason, and they are not required to achieve good reasoning performance. One could train the same model without these delimiters and likely reach similar benchmark performance.

Read full article ↗excerpt · fair-use quotation

✳ think tokens

Around this claim

Mechanism · 1

During DeepSeek-R1's RLVR training, the reasoning trace itself was not used as a training signal — the reward was based only on the final answer's correctness and format, since including the trace in training was reported to not help.Sebastian Raschka · Ahead of AI · conf 65%

Counterpoint · 1

Reasoning models can allocate compute more efficiently by using tools for tasks where they lack comparative advantage, dramatically improving test-time scaling.Brandon McKenzie and Eric Mitchell · No Priors · conf 60%

This moment responds to

extends → A 'reasoning model' is one that outputs an intermediate reasoning trace, working through a task step by step — it does not literally reason the way humans do, just as artificial neural networks don't literally work like biological ones.Sebastian Raschka · Ahead of AI