Search

Yash Sarrof

Hidden Reasoning Must Leak, but Need Not Be Readable: Fundamental Opportunities and Limits for Chain-of-Thought Monitoring
Barriers to Universal Reasoning with Transformers (and How to Overcome them)
On the Ability of Transformers to Verify Plans
Born a Transformer -- Always a Transformer? On the Effect of Pretraining on Architectural Abilities
A Formal Framework for Understanding Length Generalization in Transformers
The Expressive Capacity of State Space Models: A Formal Language Perspective

Published with Wowchemy — the free, open source website builder that empowers creators.