A Structured LLM Framework for Inorganic Material Synthesis Planning

Heewoong Noh, Gyoung S. Na, Namkyeong Lee, Chanyoung Park

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

scTrilemma: Balancing Identity, Invariance, and Reconstruction in Single-Cell Representation Learning

Yunhak Oh, Yoonho Lee, Junseok Lee, Namkyeong Lee and 8 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

How I learned to stop worrying and love StopGrads: Stationarity, Convergence, and a case study on Flow Map Learning

A stopgrad regression principle characterizes stationary points of stopgrad objectives and proves convergence to true flow maps while halving training memory.

Mark Goldstein, Max Shen, Zichu Wang, Aahlad Manas Puli and 1 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

When Riemann flows with Wasserstein: Generative Modeling of Probability Distributions on Manifolds

RWEFM generatively models meta-distributions on manifolds via Riemannian Wasserstein flow matching, yielding valid flows and efficient GPU-optimal transport approximations for non-Euclidean data.

Doron Haviv, Edward De Brouwer, Rishabh Anand, Rex Ying and 2 more

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

When to Align, When to Predict: A Phase Diagram for Multimodal Learning

Under structured cross-modal nuisance correlation, cross-modal alignment and prediction have complementary failure modes partitioned by separation ratios into four regimes, with a data-driven procedure identifying preferred objectives and when neither beats single-modality baselines.

Ilay Kamai, Hugues Van Assel, Aviv Regev, Hagai B Perets and 1 more

Paris Poster Session 5, Fri, Dec 11, 11:30 AM–1:30 PM, Paris Poster Hall · Published 2026

Generate in Reconstruction Space, Match in Semantic Space: Transport Geometry for One-Step Generation

Matching in self-supervised feature spaces improves one-step generation via compact geometry, reducing ImageNet FID by 39x and revealing metric hacking risks.

Hugues Van Assel, Edward De Brouwer, Saeed Saremi, Gabriele Scalia and 1 more

Paris Poster Session 4, Thu, Dec 10, 5:30 PM–7:30 PM, Paris Poster Hall · Published 2026