AstraFlux Data 中文

ASTRAFLUX DATA

Train for the work.
Evaluate the result.

Task environments, agent trajectories and task outputs for AI post-training and evaluation. Built around the work your model needs to do.

Explore our data Contact us
Delivery structure 01 / 03
  1. 01
    The task

    Requirements, source material and a starting state.

  2. 02
    The environment

    Tools and inputs for the work at hand.

  3. 03
    The outputs

    Task artifacts and agent trajectories.

AI POST-TRAINING & EVALUATION

CAPABILITIES

Real tasks.
Across complex domains.

Explore data programs and task environments by the capability you want to train.

01 CAPABILITY

Software & terminal

Repository engineering and terminal tasks, from long-horizon system builds to binary parsing and recovery.

Container environments Acceptance tests
Explore this capability
02 CAPABILITY

Cybersecurity

Isolated vulnerability-research tasks and security environments.

Security environments Vulnerability research
Explore this capability
03 CAPABILITY

Desktop agents (CUA)

Native Windows and macOS desktop tasks, designed around cross-application work, visual checks and restorable environments.

Windows / macOS GUI & state checks
Explore this capability
04 CAPABILITY

Professional workflows

Business APIs, document deliverables and desktop tasks with task-specific output and process checks.

Business deliverables Tools & desktop
Explore this capability
05 CAPABILITY

Deep research

Literature-grounded STEM questions with reference reasoning, answers and itemized grading criteria.

Research QA Grading rubrics
Explore this capability
06 CAPABILITY

Finance & quant

Multi-turn financial research and quantitative prediction tasks with explicit inputs and scoring contracts.

Financial research Quantitative tasks
Explore this capability
07 CAPABILITY

AI for science

Scientific analysis, model calibration and experimental decisions, with research inputs and executable checks.

Scientific computing Modeling & experiments
Explore this capability
08 CAPABILITY

Visual QA (VQA)

Chart-image questions for numerical reading, cross-panel matching and graph reasoning.

Chart reasoning Structured answers
Explore this capability
09 CAPABILITY

Design workspace

Source assets, editing records, native editable files and final renders for design tasks.

Editable PSD Visual checkpoints
Explore this capability

How we work

From a target capability
to a useful dataset.

Define the scope. Review representative tasks and acceptance criteria before agreeing the delivery.

  1. 01

    Define the work

    Agree on the target capability, task types, permitted sources and expected outputs.

  2. 02

    Review representative tasks

    Inspect representative tasks, environment behavior, scoring and model trajectories.

  3. 03

    Agree the delivery

    Set the accepted task versions, review criteria, data format, volume and timeline.

WORK WITH ASTRAFLUX

What should your
model be able to do?

Share the capability, task types and delivery requirements. We’ll discuss the right program and representative tasks.

Contact us norman383@astraflux.space