Project summary

I am a second-year B.Tech CSE student at Manipal University Jaipur, India, and the first author of “Metric-as-Feature Leakage: A Feature-Side Failure Mode in Retriever Routing,” which has been accepted as a poster at the NewInML Workshop at NeurIPS 2026 in Paris.

This is my first research paper and my first opportunity to present my research internationally. My work studies how information derived from evaluation outcomes can leak into the features used by retriever-routing systems, making a system appear better than it really is.

I want to attend the workshop in person to present the work, discuss it with researchers, and use that feedback to improve my next research projects. The main barrier is the cost of travelling from India to Paris.

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

The first goal is to present my accepted research at the NewInML Workshop at NeurIPS 2026.

The second is to get direct feedback from researchers working on retrieval, evaluation, and AI systems. I want to use those discussions to identify weaknesses in the current work and decide what experiments should come next.

The third is to continue building reproducible research around reliable evaluation of AI systems and share the work and research artifacts publicly after the conference.

How will this funding be used?

The funding would be used for the remaining costs of attending NeurIPS 2026 in Paris, primarily international airfare, accommodation, visa and travel documentation, local transportation, food, and other necessary conference-related expenses.

I am applying to several other funding sources as well. If I receive support from another source, I will update the remaining funding requirement accordingly rather than treating this as duplicate funding.

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

I am the primary person responsible for this project and the first author of the accepted paper. The research involved controlled ablations, multi-seed experiments, statistical evaluation, cross-dataset analysis, and reproducibility work.

This is my first published/accepted research project. I am continuing to work on reliable AI evaluation, retrieval systems, reinforcement learning, and agentic AI as an undergraduate researcher.

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

The main failure mode is that I do not raise enough funding to travel to Paris. In that case, I may not be able to present the poster in person even though the paper has been accepted.

The research itself will continue regardless, and I will keep the work and research artifacts available publicly. What would be lost is the opportunity for in-person presentation, discussion, and direct feedback from the NeurIPS research community.

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

I have not raised any external funding in the last 12 months.