Picture for Michael Hahn

Michael Hahn

Saarland University

Lower Bounds for Chain-of-Thought Reasoning in Hard-Attention Transformers

Add code
Feb 04, 2025
Viaarxiv icon

Emergent Stack Representations in Modeling Counter Languages Using Transformers

Add code
Feb 03, 2025
Viaarxiv icon

A Formal Framework for Understanding Length Generalization in Transformers

Add code
Oct 03, 2024
Figure 1 for A Formal Framework for Understanding Length Generalization in Transformers
Figure 2 for A Formal Framework for Understanding Length Generalization in Transformers
Figure 3 for A Formal Framework for Understanding Length Generalization in Transformers
Figure 4 for A Formal Framework for Understanding Length Generalization in Transformers
Viaarxiv icon

Separations in the Representational Capabilities of Transformers and Recurrent Architectures

Add code
Jun 13, 2024
Viaarxiv icon

InversionView: A General-Purpose Method for Reading Information from Neural Activations

Add code
May 27, 2024
Viaarxiv icon

The Expressive Capacity of State Space Models: A Formal Language Perspective

Add code
May 27, 2024
Viaarxiv icon

Linguistic Structure from a Bottleneck on Sequential Information Processing

Add code
May 20, 2024
Viaarxiv icon

Why are Sensitive Functions Hard for Transformers?

Add code
Feb 25, 2024
Figure 1 for Why are Sensitive Functions Hard for Transformers?
Figure 2 for Why are Sensitive Functions Hard for Transformers?
Figure 3 for Why are Sensitive Functions Hard for Transformers?
Figure 4 for Why are Sensitive Functions Hard for Transformers?
Viaarxiv icon

A Cross-Linguistic Pressure for Uniform Information Density in Word Order

Add code
Jun 06, 2023
Viaarxiv icon

A Theory of Emergent In-Context Learning as Implicit Structure Induction

Add code
Mar 14, 2023
Viaarxiv icon