Skip to content

A research community at INAF

Generative AI
for your research workflow.

Explore practical approaches, assess their limitations, and share what you learn with colleagues across INAF.

From noise to structure, as in a diffusion model · Try it yourself →

Gen.IA is a bottom-up initiative started by colleagues within CSN5, in collaboration with USC-C, to explore how generative AI can support every area of INAF’s work, and where it should not be relied on.

See our activities →

Start with a task

How could AI support your work?

Choose a task. Expand its guidance to see practical steps and checks.

01 / CODE

Coding &
data analysis

Draft functions, understand unfamiliar code, and document an analysis. Test every suggested solution.

Check AI-generated code
  1. Define a small task, its inputs, and the expected output.
  2. Use synthetic or shareable example data and describe the constraints.
  3. Review the suggested code, dependencies, and assumptions.
  4. Run tests against known results and relevant edge cases.
Check before useDoes the code produce correct results, and can you explain what it does?
Tools & Docs →
02 / LITERATURE

Scientific
literature

Summarise selected papers and organise evidence. Check each claim against the original source.

Verify claims and sources
  1. Select source texts you are permitted to share with the tool.
  2. Ask a focused question and ask the tool to quote the supporting passages.
  3. Compare the response with the original papers.
  4. Keep a record of sources, unresolved questions, and corrections.
Check before useCan you trace each claim and citation to a source you have verified?
Prompt gallery →
03 / COMMUNICATION

Writing &
communication

Revise a paragraph or adapt a research explanation. Retain control over meaning and scientific accuracy.

Review a revised text
  1. Provide an excerpt you are permitted to share and state the intended audience.
  2. Request a specific revision, such as clearer wording or a shorter explanation.
  3. Compare the original and revised text for changes in meaning.
  4. Review numbers, references, uncertainty, and terminology yourself.
Check before useDoes the revision preserve your scientific claims and their limitations?
Prompt engineering →
04 / EXPERIMENTS

Testing a
new approach

Compare an AI-assisted method with your existing process. Record accuracy, effort, and the ways each method fails.

Compare methods fairly
  1. Choose a representative task and define success before testing.
  2. Establish a baseline using your existing process.
  3. Compare outputs using the same criteria, including failures.
  4. Record the model or tool version, settings, effort, and costs.
Check before useDoes the approach improve the task enough to justify the work required to verify it?
Open source / open weights →

Latest highlights

AI and science: what we are reading.

All highlights →
New · interactive

Explore the AI landscape in astrophysics

A galaxy of the latest papers on generative AI, grouped by topic, and the monthly trend in astro-ph since 2022.

Open the paper galaxy →

Responsible use

Build verification into the workflow.

Protect your data

Check INAF rules and policies and the tool’s terms before sharing unpublished, personal, or restricted material.

Verify the output

Check scientific claims, citations, code, and numerical results against reliable sources or tests.

Document the process

Record relevant inputs, tool versions, settings, and validation so others can assess your approach.

Disclose AI use

Review applicable journal, funder, and institutional requirements for disclosure and attribution.

Training & shared resources

Learn together: seminars and workshops.

Gen.IA promotes the exchange of knowledge through seminars and workshops. Slides and recordings of past events are collected with the course materials.

Courses and materials →

Suggest a workshop topic

Describe a task, the tools you are considering, and what you would like to learn.

Share a tutorial or recording

Help colleagues reuse your work by including the intended audience, prerequisites, and language.

The Gen.IA community

Share what worked.
And what didn’t.

Tested a tool or explored a research workflow? Describe the task, what you tried, how you checked the results, and what did not work. You can also suggest a task to investigate together.

Contact Gen.IA