osmosMind

Data/Research

Research-Grade Scientific Reasoning

Reasoning data at the level of working researchers. Problems demand long, exact derivations — anomaly coefficients, curved-spacetime expansions, mechanism analysis — and each ships with a verifiable final answer plus the chain that reaches it. This is where retrieval fails and genuine reasoning shows.

Coverage

  • High-energy theory
  • General relativity
  • Chemistry
  • Biology
  • Scientific QA
  • Long-chain derivation

Deliverables

  • Problem statements
  • Reference answers
  • Derivation chains
  • Domain & knowledge-point tags
  • Difficulty tiers & sampling records

The pipeline

From source to acceptance

We don't hand-label a pile and ship it. Every category moves through a closed, instrumented loop — generated to a brief, checked by machines, adjudicated by experts, and traceable end to end — but the path each data type takes is its own.

  1. 01

    Frontier Authoring

    Working researchers author problems at the genuine edge of their field, engineering long derivation chains that resist retrieval and force the reasoning to be constructed rather than recalled.

  2. 02

    Reference Derivation

    Each item is paired with a closed-form reference answer and a fully worked derivation that flags every load-bearing step, so the solution is verifiable end to end.

  3. 03

    Independent Re-Derivation

    A second specialist re-derives the result from the statement alone, blind to the first solution, and discrepancies are adjudicated until the reference answer is provably sound.

  4. 04

    Knowledge-Point Tagging

    Every problem is annotated with its domain, sub-field, and the specific techniques it invokes, yielding a structured map of exactly which competencies each item probes.

  5. 05

    Frontier Calibration

    Difficulty is calibrated by measuring how the strongest available models fare and having domain specialists review the borderline results, positioning each item precisely along the curve from tractable to genuinely open.

  6. 06

    Stratified Sampling

    A final stratified pass holds the distribution across domains and tiers consistent, guarding the standard against drift as the collection scales.

Every run emits a learning signal that feeds back into the source set — the pipeline tightens itself, batch over batch.

A specimen

See the data itself

One real, trimmed sample from this category — the scenarios it serves, why it matters for training, and the shape of the data as delivered.

Where it’s used

  • Training long-horizon symbolic and mathematical reasoning
  • Evaluating frontier models on research-level science
  • Building verifier / reward signals from exact final answers

Why it matters for training

Critical

The scarcest reasoning data: at postdoc difficulty the ceiling is set by the annotator, so expert authorship is the whole product.

Notable features

  • PhD/postdoc difficulty
  • Closed-form verifiable answers
  • LaTeX-native statements
  • Knowledge-point tagged
physics-postdoc-level.jsonl · high-energy field theory
Domain

high_energy_field_theory_anomaly

Knowledge point
In the Abelian anomaly theory with external fields, how to redistribute the gauge anomaly among different current definitions using an allowed local cubic counterterm, and uniquely fix its coefficient via the charge gauge symmetry that must be preserved.
Problem
Let be a closed four-manifold, with all gauge fields Abelian one-forms. The gauge variation of some one-loop effective action, in the background of external fields , is Define and add the local term Choose so that, after adding , the gauge anomaly remains only where physically allowed, while the preserved charge gauge symmetry is strictly anomaly-free. Write down the expression for in terms of directly.
Reveal answer
Answer

Need a sample or a custom build?

Tell us your spec and scale — we deliver to order.

Get in Touch
← Back to Data