You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
$2,450 left to reach the funding goal.
I am Bhavya Shah, from Mumbai. I am excited to share that I have finished a Computer Engineering degree at Sardar Patel Institute of Technology in May 2026. I am working remotely as a research assistant at Georgia Techs Financial Services Innovation Lab. A paper for which I am a co-author has been accepted at the ATTRIB workshop at NeurIPS 2026 in Sydney and the ATTRIB workshop is about data attribution, which means figuring out which training documents are responsible for what a model ends up able to do. My part of the paper is the appendix on how to build the labels for that: where categories should come from what to do with documents that do not fit any of them and what checks need to happen before a label set is frozen and run over a hundred thousand documents. I am asking for help with the cost of the trip to the ATTRIB workshop. I should be upfront that I am not the author and the paper gets presented whether I am there or not. What I want out of going to the ATTRIB workshop is to meet the people doing this work because there are not many of them and most will be, in the room that week. I have just graduated,. I have no university funding and the lab the paper comes out of has told everyone it cannot pay for anyones travel.
Get to the ATTRIB workshop. Talk to people who work on data attribution.
Three things specifically:
Be there for our paper. Answer questions about the labelling section since labelling is the part I wrote.
Get feedback on the problem I cannot solve. My section argues you cannot validate a taxonomy on model agreement because several models can agree with each other and still all be wrong about a badly drawn category. What I propose instead is annotation on a small sample, with a double-annotated subset so you can measure how well humans agree with each other. That is expensive. I want to hear from people who have actually done it at scale.
Keep working on it. The group is taking steps and labelling layer is the piece I want to stay on.
It's a trip, not a project so I can't give you a number that means much.
What I'd count as it having been worth it: coming back with feedback that actually changes how we build the labelling layer and knowing people in this area that I don't know now. I'll write up what came of it. Post it publicly so theres something to check against.
NeurIPS registration, non-student rate: 450$
Return flight, Mumbai to Sydney : 900$
Accommodation, 4 nights shared/hostel: 700$
Local transport and meals: 300$
I have not received any other funding for this. NeurIPS financial assistance, Citadel and QRT all closed before the acceptance timeline made it worth applying.
Just me for this request.
What I've done:
Capability Provenance Should Steer Pretraining. Accepted at the NeurIPS 2026 ATTRIB workshop. Co-author. I wrote the appendix on building subdomain labels at corpus scale.
Multimodal Ordinal Regression with Bio+ClinicalBERT for Diabetes Risk Stratification in EHRs. Presented at ICCCMLA 2025.
Tutor-Personalized Automated Essay Scoring with Knowledge Graph-Guided Interpretability. Accepted at ICMI 2026.
Hybrid BiLSTM-Transformer for Load Forecasting and Anomaly Detection in Smart Grids. Under review at Electric Power Systems Research.
Research assistant at Georgia Tech FSIL auditing how reproducible NLP papers from ACL, NAACL and EACL between 2017 and 2025 actually are and testing whether AI agents can replicate them as well as people can.
Research intern at IIT Bombay Jan to Jul 2025. Built a sequence model for synthesising acoustic well logs from sparse data and a decision-support tool for ranking sites, for carbon capture.
GitHub: https://github.com/Bhavyashah20
LinkedIn: https://www.linkedin.com/in/bhavya-shah-30928b1a7/
Portfolio: bhavyashah09.tech
The visa is more likely to stop me than the money.
I am applying from India. Do not yet have an Australian visa. Home Affairs says that half of tourist visas are decided within 21 days and ninety percent within 33 days. I have seven weeks. The refusal rate, for applicants is around a quarter. So I am putting the visa application in now of waiting to hear about funding, which leaves room for a second try if the first visa gets refused.
If the visa does not come through or if the money does not reach the minimum I will not go. The paper is fine either way. The paper stays in the proceedings. One of the other authors presents the paper. The only thing lost is my presence, which's why I kept the amount small.
If I raise the money and then cannot travel I will give the money back.
Nothing. This is my first funding request.
There are no bids on this project.