Published in arXiv preprint, 2026
A category-theoretic account of symmetries in formal theorem proving, introducing proof equivariance and success invariance, with test-time rewriting aggregation to improve robustness.
Published in The Fourth Learning on Graphs Conference (LoG 2025), 2025
AnyCQ: a neuro-symbolic search framework guided by a GNN that classifies and retrieves answers to arbitrary conjunctive queries over incomplete knowledge graphs.
Published in 26th Computer Vision Winter Workshop (CVWW 2023), 2022
Comparison of in-the-wild text recognition methods with document OCR on structured documents; state-of-the-art in-the-wild detectors perform competitively for documents.
Published in arXiv preprint, 2026
An autonomous LLM-based data scientist for relational learning that searches SQL feature programs and models, then runs deterministic, interpretable inference without further LLM calls.
Published in The Fourteenth International Conference on Learning Representations, 2026
Relying on random walks to traverse the input graph, Flock introduces probabilistic node-relation equivariance and achieves state-of-the-art zero-shot link prediction across diverse KGs.