Gabriele Steidl

Research focus

Architecture Design (Transformers, SSMs, MoE) (2)Natural Language Processing (2)Training Efficiency & Optimization (1)Reasoning & Chain-of-Thought (1)

Frequent co-authors

Viktor Stein (1)Wuchen Li (1)Gregor Kornhardt (1)Jannis Chemseddine (1)

Papers (2)

Mar 17, 2026

2w ago·also University of South Carolina

SympFormer: Accelerated attention blocks via Inertial Dynamics on Density Manifolds

SympFormer achieves faster convergence in attention blocks by drawing inspiration from inertial Nesterov acceleration, offering a potential speedup without additional computational cost.

Viktor Stein, Wuchen Li, Gabriele Steidl

Architecture Design (Transformers, SSMs, MoE)Natural Language Processing Training Efficiency & Optimization

2w ago

Self-Aware Markov Models for Discrete Reasoning

Forget fixed schedules: this new discrete diffusion model learns when to stop, adapting computation to the complexity of each reasoning problem.

Gregor Kornhardt, Jannis Chemseddine, Christian Wald +1

Architecture Design (Transformers, SSMs, MoE)Natural Language Processing Reasoning & Chain-of-Thought

Search

Gabriele Steidl

Publication activitypapers/week, last 8 weeks

Research focus

Frequent co-authors

Papers (2)