Vansh Ramani.

Fast search starts with doing less work. PANORAMA uses spectral decay to evaluate distances incrementally and prune candidates early. The July 2026 revision reports up to 28.9× end-to-end speedups and integration into FAISS.

Vansh Ramani, Alexis Schlomer, Akash Nayar, Sayan Ranu, Jignesh M. Patel, Panagiotis Karras

An interpretable and fast framework for predicting solubility in aqueous and organic solvents.

Vansh Ramani, A. Arora, D. Kuchhal, Sayan Ranu, Tarak Karmakar

A position paper arguing for graph condensation methods that move beyond full-dataset training and model dependence.

Mridul Gupta, Samyak Jain, Vansh Ramani, Hariprasad Kodamana, Sayan Ranu

Investigating the stability of large language model reasoning: how reliably do reasoning systems behave when their inputs change?

Vansh Ramani, N. Potamitis, H. A. Arora, D. Kuchhal, L. H. Klein, A. Arora

Graph distillation without the expensive full-dataset training loop. Representative computation trees capture the structure that matters for node classification.

Mridul Gupta, Samyak Jain, Vansh Ramani, Hariprasad Kodamana, Sayan Ranu

A multi-solvent challenge and benchmark for evaluating solubility prediction beyond a single solvent.

Vansh Ramani, H. A. Arora, D. Kuchhal, S. Tatarin, L. Krasnov, Sayan Ranu, Tarak Karmakar

Modeling solute–solvent interactions with a merged-molecule graph representation, rather than treating each molecule in isolation.

Vansh Ramani, Tarak Karmakar