Topics
The papers, grouped by what they are about.
Graph networks that attach a vector space to every node and edge and learn the maps between them, to handle heterophily and oversmoothing.
- Topological Deep Learning Challenge 2026 TAG-DS 2026, First place, Track 1 (GNNs)
- Implicit Sheaf Operators for Cellular Sheaf Learning Preprint
Learning on higher-order domains such as simplicial and cell complexes, hypergraphs, and sheaves.
- Topological Deep Learning Challenge 2026 TAG-DS 2026, First place, Track 1 (GNNs)
- Implicit Sheaf Operators for Cellular Sheaf Learning Preprint
- Topological Deep Learning Challenge 2025 TAG-DS 2025
Recognizing, generating, and inspecting images.
- A Metacognitive Framework for Test-Time Adaptation METACOG-26, Top 3 paper
- A Metacognitive Perspective on Anomaly Detection METACOG-26
- Implicit Inversion turns CLIP into a Decoder ICLR 2026, Poster
Models that monitor and regulate their own processing, as studied in cognitive psychology.
- A Metacognitive Framework for Test-Time Adaptation METACOG-26, Top 3 paper
- A Metacognitive Perspective on Anomaly Detection METACOG-26
Adapting a deployed model to the data it meets after training, without labels.
- A Metacognitive Framework for Test-Time Adaptation METACOG-26, Top 3 paper
Finding the rare samples or regions that depart from a model of normal data.
Datasets, tasks, and protocols for measuring and comparing models.
- Topological Deep Learning Challenge 2026 TAG-DS 2026, First place, Track 1 (GNNs)
- Topological Deep Learning Challenge 2025 TAG-DS 2025
- AutoBench: Automating LLM Evaluation Preprint