CATEGORY // REASONING

Multi-step reasoning, expert decomposition, and adversarial samples.

Overview Production
CATEGORYReasoningKEY METRIC10 million entries

01 / DATASET OVERVIEW

MULTIMODAL DATA

SCOPE & PURPOSE

The training task.

  • MULTI-STEP REASONING
  • DOMAIN EXPERTS
  • ADVERSARIAL SAMPLES

Collects problems requiring multi-step reasoning (causal judgment, logic puzzles); experts decompose them into minimal reasoning units, build reasoning-flow diagrams, design adversarial samples to catch logical traps, and produce multiple solution versions.

02 / DATASET PROFILE & TECHNICAL SIGNALS

SCOPE · SCALE · CONSTRUCTION
01KEY METRIC
10 million entries
02PROBLEM DESIGN
Multi-step
Causal judgment and logic puzzles.
03REASONING STRUCTURE
Minimal units
Expert-built reasoning-flow diagrams.
04SOLUTION COVERAGE
Multiple versions
Different solution versions for each problem.

03 / DATA PRODUCTION & CONSTRUCTION

METHOD & WORKFLOW
  1. 01COLLECT

    Multi-step problems

    Collect problems requiring multi-step reasoning, including causal judgment and logic puzzles.

  2. 02DECOMPOSE

    Minimal reasoning units

    Experts decompose problems into minimal reasoning units.

  3. 03MAP

    Reasoning-flow diagrams

    Build diagrams that make the structure of the reasoning explicit.

  4. 04CHALLENGE

    Adversarial samples

    Design samples to catch logical traps and produce multiple solution versions.

04 / COVERAGE & CAPABILITY MATRIX

TASKS · METHODS · APPLICATIONS
DIMENSIONCOVERAGE
01Problem categories
Causal judgmentLogic puzzles
02Reasoning structure
Minimal reasoning unitsReasoning-flow diagrams
03Challenge coverage
Adversarial samplesLogical traps
04Solution diversity
Multiple solution versions
Explore the production method

NEXT STEP