Constrained MDPs with Trajectory Constraints

Martino Bernasconi, Matteo Castiglioni, Alberto Marchesi

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

The Benefits of Temporal Correlations: SGD Efficiently Learns k-Juntas from Random Walks

Elisabetta Cornacchia, Dan Mikulincer, Elchanan Mossel

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

Large Language Model Failures from Hallucination to Homogenization Are Different Facets of Miscalibration

Tiancheng Hu, Caiqi Zhang, Dirk Hovy, Nigel Collier

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Emergence of Distortions in High-Dimensional Guided Diffusion Models

Classifier-free guidance induces mismatched sampling distributions in diffusion models, with high-dimensional Gaussian distortions emerging when class counts scale exponentially with dimension, and a negative-guidance schedule improves diversity and separability.

Enrico Ventura, Beatrice Achilli, Luca Ambrogioni, Carlo Lucibello

Paris Poster Session 1, Wed, Dec 9, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

The Interplay of Data Structure and Imbalance in the Learning Dynamics of Diffusion Models

Diffusion models learn high-variance classes first, but sampling imbalance can reverse this ordering and delay minority class learning, causing uneven memorization.

Flavio Nicoletti, Chenxiao Ma, Enrico Ventura, Luca Saglietti and 1 more

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

Hyperbolic Graph Neural Networks Under the Microscope: The Role of Geometry–Task Alignment

Hyperbolic graph neural networks outperform Euclidean models only when tasks align with hyperbolic geometry, not merely when graphs are tree-like.

Dionisia Naddeo, Jonas Linkerhägner, Nicola Toschi, Geri Skenderi and 1 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026