Optimal In-Context Learning of Autoregressive Processes under Heterogeneous Second-Order Moments of the Prompts

Hanna Tseran, Masaaki Imaizumi

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

CITE: Anytime Valid Statistical Inference in LLM Self-Consistency

CITE provides anytime-valid certification of a target answer as the unique mode under arbitrary data-dependent stopping without knowing the answer set, with optimal stopping-time rates and improved LLM self-consistency.

Hirofumi Ota, Naoto Iwase, Yuki Ichihara, Junpei Komiyama and 1 more

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

Minimax Optimal Estimation of Transport-Growth Pairs in Unbalanced Optimal Transport

This paper develops minimax-optimal estimators for transport-growth pairs in unbalanced optimal transport and proves matching lower bounds via a stability reduction.

Donlapark Ponnoprat, Noboru Isobe, Masaaki Imaizumi

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

Training-Induced Escape from Token Clustering in a Mean-Field Formulation of Transformers

Training a linear FFN in mean-field transformers drives token distributions to escape attention-induced clustering near final layers.

Noboru Isobe, Daisuke Inoue, Masaaki Imaizumi

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

Extended Wasserstein-GAN Approach to Causal Distribution Learning: Density-Free Estimation and Minimax Optimality

GANICE minimizes averaged Wasserstein risk for conditional interventional distributions using an extended distance and cellwise critic, achieving minimax optimality without density estimation.

Shu Tamano, Masaaki Imaizumi

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Spectrum-Adaptive Generalization Bounds for Trained Deep Transformers

Spectrum-adaptive post hoc bounds for deep Transformers use layerwise Schatten quantities to trade spectral complexity against depth and hidden dimension based on learned singular-value profiles.

Mana Sakai, Masaaki Imaizumi

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 2 on Hugging Face