Yi Li on nodal superconductivity
From Young Research Leaders in Topological Materials and Beyond - Yi Li · · Flatiron Institute
“We define fundamental nodes that contribute directly to the total vorticity on a Fermi surface; additional ±1 nodes can be added in pairs, but which vortex pattern is preferred depends on the microscopic pairing mechanism.”
On , Yi Li, CEO & Co-Founder at Together AI, spoke about nodal superconductivity during Young Research Leaders in Topological Materials and Beyond - Yi Li on Flatiron Institute.
About Yi Li
Yi Li, CEO and Co-Founder of Together AI, presented research at ICCV 2025 on adapting foundational segmentation models for domains with limited labeled data. Li proposed a domain adaptation strategy that augments input images rather than fine-tuning the model itself, using a heterogeneous searching space of 10 learning-based and 22 rule-based methods. The approach employs reinforcement learning with Proximal Policy Optimization to select optimal augmentation policies, and Li reported that the method outperformed fine-tuning in common and camouflage domains in 5-shot and 10-shot experiments, though improvements were limited in ultrasonic and industrial domains. Li has also presented work on topological materials and cancer detection. In a 2024 talk on topological materials, Li discussed monopole superconductivity in Weyl semimetals, describing how monopole harmonic gap functions shift partial-wave decomposition to higher channels and how the topology threads through all energy scales. In a 2018 presentation at MIDL, Li introduced the Neural Conditional Random Field (N-CRF) algorithm for cancer metastasis detection, which models spatial correlation between neighboring image patches to improve classification accuracy. Li reported that on the Camelyon16 dataset, the method achieved an average FROC improvement from approximately 0.72 to 0.81 compared to baseline methods.