Project summary

My paper "Beyond Pattern Matching: Bridging the Sim-to-Real Gap in Transient Flare Verification Using Vision-Language Models" (https://openreview.net/forum?id=vgzXKpTZgA) was accepted at the NeurIPS 2026 Workshop on Sim2Science: ML with Imperfect Scientific Models (120 of 283 submissions accepted). The workshop is in person only, in Paris on December 12. I have registered with NeurIPS financial assistance, which covers my registration and hotel. I am asking for help with the remaining costs.

What are this project's goals? How will you achieve them?

To present the paper in person, get feedback on it, and meet researchers working on machine learning with imperfect scientific simulators, including faculty at programs I am applying to for my PhD. The paper tests whether vision-language models' judgements on real scientific data reflect reasoning or surface pattern-matching learned from simulations, a question about how far we can trust these models outside the data they were developed on.

How will you know it worked?

I present the poster on December 12, get feedback I can use in the paper's next version, and talk with researchers in the area, including faculty at programs I am applying to. I will post a short update afterwards with what came out of it.

How will this funding be used?

The funding covers lean workshop travel costs. NeurIPS financial assistance already covers my registration and hotel.

- $1,100: round-trip economy flight (India to Paris)

- $130: Schengen visa and application centre fees

- $40: travel insurance (required for the visa)

- $50: poster printing

- $280: meals and local transport (4 days)

The minimum ($500) covers everything except the flight. I have also asked the workshop organisers and sponsors for help with the flight. If the flight is covered elsewhere, I will return the flight portion of any donations. If the total raised is not enough for me to attend, I will return the donations.

Who is on your team? What's your track record on similar projects?

I am an independent researcher with an MS in physics. I work with a research group at the University of Minnesota as an external collaborator, not as a student or employee, so I have no institutional travel funding. I am the first author of the FLARE paper on detecting flares in quasar light curves (https://arxiv.org/abs/2604.08196) and the sole author of this Sim2Science paper and a second accepted workshop paper at Machine Learning and the Physical Sciences (ML4PS 2026).

LinkedIn: https://www.linkedin.com/in/atal-agrawal-735b89218/

What are the most likely causes and outcomes if this project fails?

If not enough is raised, I will not be able to attend. The paper remains public on OpenReview, but I lose the chance to present it, get feedback, and meet a potential PhD supervisor who is speaking at the workshop.

How much money have you raised in the last 12 months, and from where?

NeurIPS 2026 financial assistance (registration and hotel, in kind). No other funding.