Scholar
Antonio Vergari
Google Scholar ID: YK0NLaUAAAAJ
Reader (Associate Professor), University of Edinburgh, UK
Citations & Impact
Citations
2,385
H-index
23
i10-index
46
Publications
20
Co-authors
102
list available
Contact
Publications
Resume
Academic Achievements
- Paper highlighting that complex query answering benchmarks are far from truly complex accepted as a Spotlight (top 2.6%) at ICML 2025.
- Work on improving logical consistency of LLMs accepted at ICLR 2025.
- Proposed building deep subtractive mixture models via squaring circuits; published as ICLR 2024 Spotlight (top 5%).
- Reinterpreted KGE models (e.g., CP, RESCAL) as circuits to enable generative capabilities and guarantee logical constraints; NeurIPS 2023 Oral (top 0.6%).
- Designed a differentiable layer for neural networks to ensure predictions satisfy symbolic constraints; NeurIPS 2022.
- Developed a systematic framework for composable primitives in tractable inference; NeurIPS 2021 Oral (top 0.6%).
- Developed the cirkit library for building, learning, and reasoning with probabilistic circuits and tensor networks.
