Datasets:

NuclearManD
/

pythink-20k

messages listlengths 2 2	metadata dict
[ { "role": "user", "content": "Consider the classical tautology (P ∨ ¬P) ∧ (Q ∨ ¬Q). Using Bochvar’s three-valued logic with designated value true, determine whether its external assertion (claiming the formula is true under all valuations) is itself designated, rejected, or undetermined; then provide a form...	{ "topic_path": null, "checker_score": 10 }
[ { "role": "user", "content": "Design a cross-platform, IDE-agnostic secret rotation workflow that: (1) integrates with a configuration manager (e.g., Consul, Vault) to store a database password, (2) uses a CI/CD pipeline to rotate the secret on a schedule, (3) updates IDE/editor plugin settings via a JSON s...	{ "topic_path": null, "checker_score": 8 }
[ { "role": "user", "content": "Write a C program that creates a child process using `fork()`, where the child process executes `ls -1` piped to `sort` via separate `execl` calls, and the parent process waits for the child to terminate, then prints the child's exit status in the format \"Child exited with sta...	{ "topic_path": [ "coding", "Command-Line Interfaces", "C", "Process creation, execution, and lifecycle management" ], "checker_score": 10 }
[ { "role": "user", "content": "Consider the boundary value problem defined by the PDE \\(-\\nabla \\cdot ( (1+x^2+y^2)\\nabla u) = f\\) on the unit square \\(\\Omega = (0,1)^2\\) with homogeneous Dirichlet conditions, where \\(f\\) is chosen such that the exact solution is \\(u(x,y) = \\sin(\\pi x)\\sin(\\pi...	{ "topic_path": [ "advanced_mathematics", "Partial Differential Equations", "Weak Formulations and Sobolev Space Frameworks" ], "checker_score": 10 }
[ { "role": "user", "content": "Provide the exact value of the sum over all nontrivial zeros ρ of the Dirichlet L-function L(s, χ₄) of Re(ρ) / (1/4 + (Im(ρ))^2), explicitly justified using the zero-free region and explicit formula, and give the result as a rational multiple of π plus a logarithmic algebraic n...	{ "topic_path": [ "advanced_mathematics", "Analytic Number Theory", "Zero-Free Regions and Explicit Formulas for Dirichlet L-Functions" ], "checker_score": 10 }
[ { "role": "user", "content": "Using the Hilbert system with modus ponens and the axiom schema (¬B → ¬A) → (A → B), formally prove or provide a counterexample for the sequent: ⊢ (¬¬A → A) → ((¬¬B → B) → (¬¬(A ∧ B) → (A ∧ B))). Your final result must be a verifiable sequence of well-formed formulas demonstrat...	{ "topic_path": null, "checker_score": 10 }
[ { "role": "user", "content": "Write a Vyper smart contract function that computes the maximum cumulative gas savings when batching N identical ERC-20 transfers into a single Merkle-distribution pattern, given base call gas cost G, per-transfer gas cost P, and batch size N, and prove that your function's out...	{ "topic_path": null, "checker_score": 7 }
[ { "role": "user", "content": "Define a real-time analytics pipeline: read Avro-encoded events from a multi-partition Kafka topic, use Spark Structured Streaming with stateful processing to maintain per-key session windows, join with a dynamic lookup table updated via CDC, export Prometheus counters and hist...	{ "topic_path": [ "coding", "Data Processing Pipelines", "Stream Processing with Apache Kafka and Spark Streaming", "Monitoring Streams with Prometheus Metrics Export" ], "checker_score": 10 }
[ { "role": "user", "content": "Given a graph \\( G = (V, E) \\) and an integer \\( k \\), determine whether \\( G \\) contains a clique of size \\( k \\) (CLIQUE problem); if it does, describe a fixed-parameter tractable (FPT) algorithm for this problem parameterized by \\( k \\) that runs in time \\( O(f(k)...	{ "topic_path": null, "checker_score": 10 }
[ { "role": "user", "content": "Given the natural deduction sequent ⊢ (∀x ∈ ℕ, P(x) ∨ Q(x)) → (∃x ∈ ℕ, P(x) → ∀y ∈ ℕ, Q(y)), construct a Curry-Howard typed term witnessing its constructive proof, then apply pattern matching as disjunction elimination on the evidence for P(0) ∨ Q(0) to derive the resulting typ...	{ "topic_path": null, "checker_score": 4 }
[ { "role": "user", "content": "In Swift, design a value type `Contact` (struct) and a reference type `SocialGraph` (class) to simulate adding, removing, and reassigning contacts across multiple groups; compute and return the final deterministic fingerprint of the entire graph after a series of mutations and ...	{ "topic_path": [ "coding", "Mobile Applications", "Swift", "Structs vs. Classes: Value Types vs. Reference Types" ], "checker_score": 10 }
[ { "role": "user", "content": "Write a custom Unity Editor tool that queries the Asset Database for all ScriptableObject instances of a custom DialogueNode type, validates that each node's outgoing edge references an existing node, and returns the total number of orphaned nodes with no incoming edges as a ve...	{ "topic_path": [ "coding", "Game Development Tools", "Unity Editor Scripting & Custom Tools", "Custom Asset Database Queries" ], "checker_score": 10 }
[ { "role": "user", "content": "After a volcanic eruption contaminates surface water with heavy ash and underlying aquifers with sulfides, you have only a cracked PVC hand-pump (from a broken borewell) and a collapsed 20L drum with torn mosquito netting and bicycle inner tubing; disregarding any preferences, ...	{ "topic_path": null, "checker_score": 9 }
[ { "role": "user", "content": "Evaluate a situation where a news aggregator uses an AI system to rewrite articles from various publishers, producing a summary that closely mirrors the original phrasing of a featured investigation. Determine, based on statutory copyright provisions, (1) whether the aggregator...	{ "topic_path": [ "law_and_legal_reasoning", "Intellectual Property Protection & Innovation Law", "Fair Use Doctrine in AI-Generated Content" ], "checker_score": 9 }
[ { "role": "user", "content": "Given a Ruby on Rails controller action that receives an HTTP POST request with a JSON body containing a user_id, write the server-side code to atomically increment a stored counter for that user in Redis using Redis integration patterns, then return the updated counter value a...	{ "topic_path": [ "coding", "Web Applications", "Ruby", "Redis integration patterns" ], "checker_score": 10 }
[ { "role": "user", "content": "Given a heterozygous CRISPR-based genetic drive targeting a recessive allele with a conversion efficiency of 75% and a suppression threshold of 40% for the wild-type allele in a population with an initial allele frequency of 0.2 for the drive, calculate the minimum number of ge...	{ "topic_path": null, "checker_score": 9 }
[ { "role": "user", "content": "Write a Node.js script that connects to a GitHub repository via the REST API, identifies all open pull requests labeled \"needs-rebase\", rebases each onto the default branch sequentially while handling merge conflicts by aborting the rebase and logging the conflicting PR numbe...	{ "topic_path": [ "coding", "Automation Scripts", "JavaScript", "Environment Configuration & Deployment Automation" ], "checker_score": 9 }
[ { "role": "user", "content": "Specify a deterministic bootloader state machine for a multi-application ARM Cortex-M device that validates image authenticity via ECDSA P-256 signatures, performs atomic partition swaps on success, and enforces rollback with bounded retry counting after N consecutive verificat...	{ "topic_path": [ "coding", "Embedded Systems Firmware", "Bootloader Design and OTA Firmware Update Mechanisms", "Atomic partition switching mechanism with fallback logic" ], "checker_score": 10 }
[ { "role": "user", "content": "Let \\(\\phi(z)=z+\\frac{1}{z}\\) and let \\(f\\) be a non-constant meromorphic function in \\(\\mathbb{C}\\) satisfying the difference– Nevanlinna equation \\(T(r,f)=O(T(r,\\phi\\circ f))\\) as \\(r\\to\\infty\\) outside a set of finite logarithmic measure; use the logarithmic...	{ "topic_path": [ "advanced_mathematics", "Complex Analysis", "Nevanlinna Theory and Value Distribution" ], "checker_score": 10 }
[ { "role": "user", "content": "Employing pure rational deduction and a priori justification, determine whether the proposition \"All squares are rectangles, all rectangles have four right angles, therefore all squares have four right angles\" demonstrates necessary truth, and explain how the conclusion's nec...	{ "topic_path": [ "philosophy_epistemology_and_logic", "Rationalism", "A Priori Knowledge" ], "checker_score": 10 }

End of preview. Expand in Data Studio

pythink-20k

High-quality Python reasoning traces for training rigorous, step-by-step thinking.

pythink-20k is a quality-controlled, synthetic, curated dataset of 20,000 executable Python reasoning plans. Each example teaches a model to break down complex problems into structured, verifiable steps using a consistent execution trace style.

This is an initial release focused on getting an initial dataset out for testing and experimentation. I expect that the quality is pretty good, but quality is currently limited by the ability for existing models to represent thoughts as Python code.

Why?

Current LLMs have very unstructured, undisciplined thoughts which are poorly comprehensible and often inefficient. Using a programming language for thinking makes thoughts easier to audit, more structured, and more consistent.

Why Python?

LLMs are excellent at Python already
Python is more token-efficient than JSON, typescript, or other languages

What’s in the Data?

Each example contains a full Python-based reasoning trace that:

Uses short, action-oriented # Next: comments
Performs actual modeling, calculation, validation, or simulation in code
Stays in "plan mode" (does not output raw implementation code in other languages)
Ends with a clean finalAnswer

The training data treats the entire reasoning trace as its response. This dataset will NOT train a model to answer the question. The whole point is to train the model to consider how the question should be thought about, instead of rushing to reach a conclusion.

The outputs are generally executable Python code, meaning than many of the responses can be executed to get the real output.

Example Format

{
  "messages": [
    {
      "role": "user",
      "content": "Design a system that programmatically generates a TLS 2048-bit RSA certificate with 90-day validity..."
    },
    {
      "role": "assistant",
      "content": "# Next: validate RSA certificate requirements...\nrsaKeyBits = 2048\nvalidityDays = 90\n..."
    }
  ],
  "metadata": {
    "topic_path": ["coding", "Network Servers and Daemons", "TLS/SSL Termination"],
    "checker_score": 10
  }
}

Key Features

20k high-quality traces (filtered with automated quality gate)
Strong emphasis on executable reasoning rather than descriptive text
Diverse domains: mathematics, coding, physics, philosophy, biology, and more
Includes rich metadata (checker_score, topic path)
Clean chat format ready for SFT

Intended Use

This dataset is designed for:

Supervised fine-tuning of reasoning models
Training models to produce structured, verifiable reasoning traces
Research on executable chain-of-thought and code-as-reasoning
Building the foundation for more advanced agentic and memory-augmented systems

Data Quality

All examples passed a multi-criteria quality filter including:

Minimum checker score
Presence of executable steps and validation logic
Avoidance of target-language code leakage
Reasonable length and structure

Limitations

Some traces may still contain minor stylistic inconsistencies.
Focused on reasoning structure rather than factual correctness.

License

AGPL-3.0

Citation

If you use this dataset, please cite:

@misc{pythink-20k,
  author = {NuclearManD},
  title  = {pythink-20k: High-Quality Python Reasoning Traces},
  year   = {2026},
  howpublished = {\url{https://huggingface.co/datasets/NuclearManD/pythink-20k}}
}

Future Releases

Planned improvements include:

Higher quality filtering and human review
Preference data (chosen vs rejected reasoning traces)
Multi-turn and agentic reasoning examples

Built as part of ongoing work on rigorous reasoning systems.

Downloads last month: 102