Prompting technique

Thinking Token-based Test-time Scaling (TTTS)

Demystifying Reasoning Dynamics with Mutual Information: Thinking Tokens are Information Peaks in LLM Reasoning, arXiv:2506.02867 (2025) · read the paper

What it does

A technique using special 'thinking tokens' (trigger words like 'Hmm', 'So', 'Wait') to stimulate deep analysis in LLMs. Embedding these words in prompts creates information peaks – moments of maximum proximity to the correct answer, significantly improving the accuracy of the model's reasoning when solving complex problems.

Task typereasoning
Shapechain of thought
Efforttwo or three steps
Addressesanswers that are close but wrong; the same prompt giving different answers; invented facts and citations
Published2025-06-03
AuthorsChen Qian, Dongrui Liu, Haochen Wen et al.

When it helps

Apply this technique when solving complex problems requiring multi-step reasoning: mathematical calculations, logical deductions, data analysis, decision-making. Particularly effective when the model gives superficial or inaccurate answers to complex questions. Use when developing prompts for critically important tasks requiring high accuracy. The method is suitable for improving reasoning quality in research work, financial analysis, strategic planning. Embed

A prompt you can paste

Generic by design: it applies the technique without knowing your task. Adapt the marked parts.

Prompt
# ROLE:
You are an AI assistant designed to enhance reasoning quality by leveraging "thinking tokens" as information peaks.

# CONTEXT:
The user needs to perform a complex task that requires deep analysis and accurate output. The specific task is to be defined by the user.

# TASK:
Perform the following task: [INSERT USER'S TASK HERE]

# KEY INSTRUCTION:
After generating the initial response, you must perform a self-critical analysis.
Begin this analysis with one of the following "thinking token" phrases:
- "Хм, давайте-ка еще раз взглянем на это..."
- "Итак, подведем итог и обозначим главные риски..."
- "Подождите, а все ли я учел при генерации ответа?"
- "Хм, а что если посмотреть на это с другой стороны..."

In this self-critical analysis section, identify at least two potential weaknesses, inaccuracies, or areas for improvement in your initial response. For each identified weakness, propose a specific, actionable way to enhance the response.

# OUTPUT FORMAT:
1.  **Initial Response:** [Your generated response to the user's task]
2.  **Self-Critical Analysis:** [Begin with one of the specified "thinking token" phrases, followed by the analysis and proposed improvements]

If this one does not fit, the two closest alternatives in the corpus are Anti-Bias Prompting and Tabular Prompting, which target the same failure from a different angle.

Worked example

The same technique applied to a concrete job: work through a multi-step planning problem. Use it as the pattern for your own case rather than as a finished artefact.

Worked example
# ROLE:

You are an AI assistant tasked with solving complex problems by leveraging the "Thinking Token-based Test-time Scaling" (TTTS) technique.

# CONTEXT:

The task is to work through a multi-step planning problem. The specific problem details are not provided, but the method requires a structured approach that simulates deep thinking and self-correction.

# TASK:

Solve the following multi-step planning problem: **[INSERT THE SPECIFIC MULTI-STEP PLANNING PROBLEM HERE]**

# KEY INSTRUCTION (TTTS Method):

After you have generated the initial plan or solution, you must perform a critical self-analysis. To do this, start a new section with the heading "**Self-Correction and Refinement**". Begin this section with the exact phrase: **"Хм, давайте-ка еще раз взглянем на этот план..."** (which translates to "Hmm, let's take another look at this plan...").

Within this "Self-Correction and Refinement" section, you must:
1. Identify at least two potential weaknesses, overlooked aspects, or areas for improvement in your initial plan.
2. For each identified weakness, propose a concrete adjustment or enhancement to the plan.
3. Ensure these adjustments align with the overall goal of the planning problem.

**The goal is to simulate a process of deep thinking, self-critique, and iterative improvement, directly applying the principles of TTTS.**

Get this written for your actual task

Paste what you are trying to do and the corpus will be matched against it directly. Free, no account, about ten seconds.

Free · no signup · ~10s
0.00match confidence
single retrieval pass
Prompt for your task

      

That number is low on purpose, and it is real. It is the raw similarity of one retrieval pass: no specialist read the paper, no judge compared anything, the first plausible match won.

6,235techniques in the corpus
one of which is this page

Picking the right one for a specific task is the work, and it is the work GetDecision does.

This pageone technique, generic prompt
What you just ranone technique matched to your wording, nothing verified
Full runten specialists read the papers in full, a judge ranks the top three for your task and shows its reasoning, generation on the model you pick, saved to your history

See the top three for your taskTen specialists, a judge, and the reasoning shown. Free account, first run included.

Run the full analysis

Related techniques

Anti-Bias PromptingThe technique of neutral prompt formulation that excludes cognitive biases and user prejudice. Essence: avoid …Tabular PromptingA method of presenting data as a markdown table before submitting a request to an LLM. The tabular structure h…Multilingual Ensemble PromptingThe method involves translating a complex query into 3-4 different languages, obtaining separate answers from …Debate-to-Detect (D2D) / Multi-Agent Structured DebateThe method transforms an analysis task into structured debates between multiple AI agents with opposite roles.…

All techniques · Failure modes and fixes